Skip to content

🦞 OpenClaw Ecosystem Digest 2026-04-16 #613

Open
Open
🦞 OpenClaw Ecosystem Digest 2026-04-16#613
Labels
openclaw-en
@github-actions

Description

@github-actions
github-actions
bot
opened on Apr 16, 2026

OpenClaw Ecosystem Digest 2026-04-16

Issues: 500 | PRs: 500 | Projects covered: 13 | Generated: 2026-04-16 00:17 UTC

OpenClaw Deep Dive

OpenClaw Project Digest — 2026-04-16

1. Today's Overview

OpenClaw remains highly active with 500 issues and 500 PRs updated in the last 24 hours, indicating a large, engaged community and rapid development pace. However, project health shows strain: a critical onboarding regression (TypeError: Cannot read properties of undefined (reading 'trim')) is blocking new user installations across multiple paths (onboard, configure, install script), with numerous duplicate reports confirming widespread impact. The team shipped one beta release focused on control UI observability, while maintainers actively closed 138 PRs/issues against 362 open PRs still in flight. The contributor base is diverse with significant community PR activity, but stability issues—particularly around installation, Windows support, and provider integrations—are dominating user attention.

2. Releases

v2026.4.15-beta.1

Aspect Detail
Type Beta
Focus Gateway observability & model authentication health

Changes:

  • Control UI/Overview: Added a Model Auth status card displaying OAuth token health and provider rate-limit pressure at a glance
  • Visual attention callouts when OAuth tokens are expiring or expired
  • Backed by new models.authStatus gateway method that strips credentials and caches for 60s

Breaking changes / migration notes: None reported.

3. Project Progress

Notable Merged/Closed PRs (2026-04-15/16)

PR Author Summary Impact
#66378 athletesofthereef Fix WhatsApp media sends by bypassing legacy outbound send dependency Channel reliability — media messaging restored
#66331 acwilan Add per-agent TTS and STT overrides Major multi-agent UX improvement; agents can now have individual voice/speech configs
#59619 MoerAI Fix Feishu TTS audio delivery as voice messages instead of file attachments Enterprise channel fix
#43961 happydog-intj SIGKILL stale processes when graceful shutdown times out Stability — prevents port conflicts on restart
#66259 seank-com Prefer --mask for QMD collection add compatibility Memory/local AI compatibility
#67395 igormf Fix Windows pnpm.exe runner and Lobster embedded runtime fallback Windows install unblocking

Features Advancing in Open PRs

PR Focus
#65554 Full media suite for WebChat (images, audio, video, documents) — major UI enhancement
#66168 MCP Apps protocol extension — tools, resources, UI meta for gateway-powered app experiences
#66898 Windows native wrapper & hardened installer
#67433 Blocking mode for POST /hooks/agent endpoint
#67420 Per-agent dreaming control to prevent OOM

4. Community Hot Topics

Most Active Issues by Engagement

Issue Comments Status Topic Underlying Need
#49971 89 Open RFC: Native Agent Identity & Trust Verification Decentralized identity, agent-to-agent trust, verifiable credentials — signals demand for inter-agent authentication standards
#45064 31 Closed Memory leak / OOM on basic CLI commands Resource efficiency in long-running or frequent CLI usage
#25592 19 Open (stale) Text between tool calls leaks to messaging channels Clean UX boundary between internal agent reasoning and user-facing output
#34830 19 Open (stale) OpenRouter 401 missing authentication header regression Provider auth reliability — regression in model routing
#35220 15 Open (stale) Codex Responses API server_error doesn't trigger fallback Resilience in model failover chains

Analysis: The community is deeply engaged in two parallel tracks: foundational trust/infrastructure (identity, agent economy, MCP protocols) and daily reliability (auth, leaks, fallbacks, memory). The 89-comment RFC on agent identity suggests OpenClaw is being positioned as infrastructure for a broader multi-agent ecosystem.

5. Bugs & Stability

Critical / Severe — Blocking New Users

Issue Severity Description Fix PR?
#66718 P0 — Critical openclaw configure crashes with TypeError: Cannot read properties of undefined (reading 'trim') Yes — #67130 closed, but reports continue
#67291 P0 — Critical openclaw onboard crashes with same trim error, cannot skip channels Yes — #67130
#67347 P0 — Critical Install quickstart crashes with trim error Same root cause
#67074 P0 New 4.14 install: trim crash on channel skip Closed as duplicate
#66641 P0 Installer crashes after "Select channel (QuickStart)" Closed as duplicate

Note: Despite #67130 being merged as "fix bundled setup plugin meta is incomplete," fresh reports on 2026-04-15 suggest the fix may be incomplete or not yet released.

High Severity — Runtime Regressions

Issue Severity Description Fix PR?
#66633 P1 — High openai-codex provider fails with Cloudflare 403 on every request after 2026.4.14 upgrade No fix PR identified
#66674 P1 — High openai-codex/gpt-5.4 direct CLI infer returns HTML rawError, surfaced as "DNS lookup failed" No fix PR identified
#66601 P1 — High v2026.4.14 causes repeated context engine errors (lossless-claw factory invalid) No fix PR identified
#67035 P1 — High Windows chat UI regression: input swallowed, streamed replies invisible, typing indicator broken No fix PR identified

Medium Severity — Functional / UX

Issue Severity Description Fix PR?
#66207 P2 — Medium Control UI chat flickers: sent messages disappear briefly then reappear No fix PR identified
#67288 P2 — Medium amazon-bedrock-mantle lacks config.discovery.enabled gate; unnecessary discovery on every request No fix PR identified
#67261 P2 — Medium Venice model responses missing id/status cause crash No fix PR identified

6. Feature Requests & Roadmap Signals

Issue/PR Request Likelihood in Next Version
#49971 Native Agent Identity & Trust Verification (ERC-8004, DID, VC) Medium — high engagement but architectural; likely spawns working group
#28106 Agent-to-Agent Task Delegation Protocol ("Agent Economy") Medium — aligned with identity RFC, needs protocol design
#28930 Memory v2: Associative Traversal, Salience Weighting, Access-Based Forgetting High — memory is a repeated pain point; closed but likely resurfacing
#34400 memory_search recursive subdirectory search (memory/**/*.md) High — small, well-scoped, clear user value
#20562 Add Serper.dev as web_search provider Medium-High — Brave Search free tier killed; community demand for alternative
#29563 Control UI session picker + "New chat" + rename Medium — UX polish with sustained interest
#66168 MCP Apps protocol extension High — large open PR, signals official direction

7. User Feedback Summary

Top Pain Points

  1. Installation/onboarding is broken — The trim crash is the loudest, most repeated complaint across languages and platforms. New users cannot get past channel selection.
  2. Provider fragility — OpenAI Codex, OpenRouter, and Venice integrations are failing after recent updates. Users experience 403s, 401s, DNS misreports, and missing fallbacks.
  3. Windows is second-class — Multiple reports of UI regressions, install failures, and WebSocket races on Windows.
  4. Leaky abstractions — Tool call text leaks to messaging channels; exec completion events pollute webchat; internal reasoning surfaces to users.

Positive Signals

  • Multi-agent features (per-agent TTS/STT) are advancing and respond to real power-user needs.
  • Media support in WebChat (#65554) is a highly anticipated community contribution.
  • Chinese i18n docs (#64605) show global adoption efforts.

Satisfaction Assessment

Dimension Score Rationale
Core stability ⚠️ At risk Onboarding regression + provider breakages in latest release
Feature velocity ✅ Strong Active PRs for media, MCP, Windows native, memory improvements
Community health ✅ Healthy High comment counts, diverse contributors, sustained engagement
Release quality ⚠️ Concerning 2026.4.14 introduced multiple regressions; beta cadence may need gating

8. Backlog Watch

Long-Unanswered Important Items Needing Maintainer Attention

Issue/PR Age Problem Why It Needs Attention
#25592 ~7 weeks Tool call text leaks to messaging channels Core UX bug; stale but unaddressed; PR #65508 partially addresses tag stripping
#34830 ~6 weeks OpenRouter 401 missing auth header Provider regression affecting a popular routing option
#35220 ~6 weeks Codex server_error doesn't trigger fallback Reliability gap in model failover; variant of previously closed issue
#43419 ~5 weeks npm install -g openclaw fails without Git Installation barrier due to libsignal-node git dependency
#42225 ~5 weeks gpt-5.4 still uses 272k context despite PR #37876 Context engineering debt; wastes tokens and degrades performance
#35071 ~6 weeks Eval framework for skill-creator Developer experience; large open PR with no recent maintainer review
#59949 ~2 weeks includedWorkDirs agent workspace support Agent workspace flexibility; XL PR needing review

Digest generated from GitHub activity data for 2026-04-16.

Cross-Ecosystem Comparison

Cross-Project Comparison Report: Open-Source AI Agent & Personal Assistant Ecosystem

April 16, 2026

1. Ecosystem Overview

The open-source AI agent ecosystem is experiencing intense, sustained development velocity across more than a dozen active projects, with daily issue and PR volumes rivaling major infrastructure software. However, the landscape is deeply fragmented — no single project has achieved dominant market position, and most are grappling with similar growing pains: onboarding regressions, provider API fragility, Windows/Linux cross-platform gaps, and the architectural leap from single-agent to multi-agent orchestration. The field is simultaneously feature-expanding (MCP adoption, voice channels, local LLM routing) and stabilization-challenged, with several projects shipping regressions that block new users. Enterprise and self-hosted deployment scenarios are emerging as the primary battlegrounds for differentiation.

2. Activity Comparison

Project Issues (24h) PRs (24h) Releases Health Score* Assessment
OpenClaw 500 500 v2026.4.15-beta.1 ⚠️ C+ Massive scale, but critical onboarding regression blocking new users; provider breakages in latest release
NanoBot 17 (11 open) 46 (28 open) None ⚠️ B Very high velocity, fast integration phase; memory regression risk from v0.1.5.post1
Hermes Agent 50 (22 open) 50 (31 open) None ✅ B+ Rapid stabilization, strong maintainer response; healthy backlog with architectural milestones closing
PicoClaw 7 (4 open) 26 (6 open) v0.2.6-nightly ✅ A- Highest merge rate, same-day maintainer turnaround; strong acceleration with mature multi-agent features
NanoClaw 5 fresh 17 (13 open) None ⚠️ B- Aggressive feature expansion but concentrated onboarding friction; zero maintainer response to new issues
NullClaw 9 8 (5 open) None ⚠️ C+ Active contributions but basic feature breakages (shell, HTTP) and source-build friction
IronClaw 17 (9 open) 50 (43 open) None ⚠️ C+ Very high velocity but P0 security regression + staging web UI regressions; backlog accumulating
LobsterAI 3 27 (5 open) None ✅ B Fast core-team throughput, limited external community; enterprise Chinese IM focus
Moltis 7 (1 open) 15 (5 open) None ✅ A- Exceptional close rate (6/7 issues resolved); reactive but effective provider patching
CoPaw/QwenPaw 50 (25 open) 50 (25 open) v1.1.2-beta.1, v1.1.1.post1 ⚠️ B- Aggressive iteration with dual releases; silent failures and branding confusion eroding trust
ZeroClaw 50 (39 open) 41 (34 open) None ⚠️ C Backlog growing faster than resolution; 7 open S0 issues, persistent onboarding blockers
TinyClaw 0 0 None — Inactive No activity
ZeptoClaw 0 0 None — Inactive No activity

*Health score synthesizes velocity, close/merge rate, severity of open bugs, and maintainer responsiveness.

3. OpenClaw's Position

Advantages vs. Peers

  • Scale dominance: OpenClaw's 500 issues and 500 PRs in 24 hours dwarfs every competitor except CoPaw/QwenPaw and ZeroClaw, reflecting the largest contributor base and most diverse use cases.
  • Protocol leadership: The 89-comment RFC on Native Agent Identity & Trust Verification (#49971) and the active MCP Apps protocol extension (#66168) position OpenClaw as the de facto standards incubator for inter-agent authentication and multi-agent tooling.
  • Channel breadth: WhatsApp, Feishu, WebChat media suite, and per-agent TTS/STT overrides demonstrate the most mature multi-channel infrastructure.

Technical Approach Differences

  • Gateway-centric architecture: OpenClaw invests heavily in a unified gateway abstraction (models.authStatus, control UI observability) rather than per-channel silos, contrasting with NanoBot's or LobsterAI's connector-heavy approaches.
  • Beta-first release cadence: Ships frequent beta releases with observability focus, whereas PicoClaw and Moltis prefer nightly/patch stability, and CoPaw pushes aggressive minor releases.

Community Size Comparison

Metric OpenClaw Nearest Comparable
Daily issues+PRs ~1,000 CoPaw: ~100, ZeroClaw: ~91
Comment velocity on RFCs 89 (#49971) Hermes multi-agent: 27, CoPaw help wanted: 57
Contributor diversity Very high (athletesofthereef, acwilan, MoerAI, seank-com, igormf, etc.) Moltis: concentrated (penso, Cstewart-HC); LobsterAI: core-team dominated

Vulnerability: OpenClaw's scale has become a liability for release quality. The 2026.4.14 release introduced multiple regressions, and the critical trim onboarding crash is generating duplicate reports faster than fixes propagate.

4. Shared Technical Focus Areas

Requirement Projects Specific Needs
MCP (Model Context Protocol) tooling OpenClaw (#66168), PicoClaw (#2535), Moltis (#712, #724, #732, #733), NanoClaw (#1781), IronClaw (#2474) Health-check robustness, stdio vs. Streamable HTTP server support, slash-command UX, OAuth discovery boundaries
Multi-agent orchestration OpenClaw (#49971, #28106), Hermes (#344), PicoClaw (#2531), CoPaw (#3340), ZeroClaw (#5774) Agent-to-agent delegation, trust verification, ACP-based external delegation, identity standards (ERC-8004, DID, VC)
Local/self-hosted LLM support NanoBot (#3186), PicoClaw (#28), CoPaw (#3443, vLLM/Ollama issues), Moltis (#723, #727), ZeroClaw (#5531) LM Studio frictionless connect, Ollama fixes, context window overrides, local/cloud routing, judge-based routing
Provider resilience & fallbacks OpenClaw (#34830, #35220, #66633), NanoBot (#3143), Hermes (#10223, #10575), Moltis (#716, #717, #725), ZeroClaw (#5670, #5527, #5600) Auth header regressions, 401/403 handling, Codex/OpenRouter failover, reasoning content preservation, strict schema compatibility
Windows-native stability OpenClaw (#66898, #67035), Hermes (#10550), ZeroClaw (#5562), NanoClaw (#1787 Apple Container) Installer hardening, UI regressions, pnpm/Node runtime alignment, WebSocket races
Memory/data reliability NanoBot (#2957, #3190), OpenClaw (#28930), Hermes (#10570, #6025), CoPaw (#3011, #3279), Moltis (#728) Dream/consolidation overwrite bugs, format compatibility, MEMORY.md reliability, context compression, unbounded state growth
Voice/TTS/STT channels OpenClaw (#66331), CoPaw (#3449), NanoClaw (#1760), Moltis (#735) Per-agent voice configs, Twilio fallback, SIP/LiveKit integration, custom ElevenLabs voices

5. Differentiation Analysis

Dimension Leaders Laggards Key Differentiators
Enterprise IM (China) LobsterAI (POPO, DingTalk, Feishu, Weixin, QQ), NanoBot (Feishu, Teams), OpenClaw (Feishu, WhatsApp) NullClaw, ZeroClaw LobsterAI's native connector depth; OpenClaw's media messaging reliability
Terminal-first / developer UX Hermes (rich output, diff views, syntax highlighting), ZeroClaw (Rust TUI) NanoBot, CoPaw Hermes' agent/rich_output.py and Web UI Gateway closure
Web UI polish PicoClaw (markdown highlighting, /btw command, composer fixes), IronClaw (dashboard, search integration) ZeroClaw (#4866 dashboard unavailable), NullClaw PicoClaw's same-day UI bug fixes; IronClaw's staging regressions are a liability
Security/sandboxing Moltis (file tool containment, approval gating), ZeroClaw (TOTP shell gating, path guards), NullClaw (sandbox PATH validation) OpenClaw, CoPaw Moltis' directory containment + approval gating; ZeroClaw's per-command TOTP
Local/edge deployment PicoClaw (Sogou search, cross-platform ARM/FreeBSD fixes), NanoBot (LM Studio first-class), CoPaw (local routing v1) OpenClaw PicoClaw's hardware-aware portability; CoPaw's judge-based local/cloud routing
Observability/control plane OpenClaw (Model Auth status card, gateway caching), CoPaw (Agent Statistics PR), IronClaw (insights interval) Hermes, NullClaw OpenClaw's 60s-cached models.authStatus gateway method
Architectural ambition ZeroClaw (microkernel RFC, crate splitting), OpenClaw (MCP Apps, identity RFC), Hermes (multi-agent architecture closed) PicoClaw, Moltis ZeroClaw's v1.0.0 microkernel restructuring; OpenClaw's protocol standardization

6. Community Momentum & Maturity

Tier 1: Rapidly Iterating, High Merge Health

Project Characteristics
PicoClaw 77% merge rate (20/26), same-day maintainer response, nightly builds, multi-agent delegation landing
Moltis 75% merge rate (15/20), 86% issue close rate, exceptional maintainer velocity (penso), provider patching excellence
Hermes Agent 38% PR merge/close rate but high-quality closures, rapid bug-fix turnaround, architectural milestones completed

Tier 2: High Velocity, Stability Strain

Project Characteristics
OpenClaw Massive throughput but release quality degrading; onboarding blocked; community scale is unmatched but management lagging
CoPaw/QwenPaw Dual releases in 24h, ambitious features (Mission Mode, routing), but silent failures and branding confusion
NanoBot Fast nightly integration, large PR splitting, but memory regression and skill-development friction
IronClaw High PR volume but backlog accumulating; P0 security + web regressions need immediate triage

Tier 3: Growing but Fragile

Project Characteristics
NanoClaw Aggressive roadmap execution but zero maintainer response to concentrated onboarding issues
NullClaw Active contribution base but basic tool breakages and build friction undermining first impressions
ZeroClaw Backlog growing faster than resolution; 7 open S0 issues; architectural RFCs may be escape hatches from reactive maintenance

Tier 4: Core-Team Dominated / Limited External Community

Project Characteristics
LobsterAI 81% merge rate but only 3 issues; NetEase-internal velocity with Chinese enterprise focus; low external contributor diversity

Inactive

| TinyClaw, ZeptoClaw | Zero activity |

7. Trend Signals

For AI Agent Developers

Trend Evidence Strategic Implication
MCP is becoming table stakes 6+ projects actively implementing or hardening MCP server/tooling support Developers should design agent skills around MCP compatibility; custom tool protocols face extinction
Agent identity and trust protocols are emerging OpenClaw's 89-comment RFC (#49971), Hermes' multi-agent architecture closure, CoPaw's ACP delegation PR Inter-agent authentication will be the next major standards war; early adoption of DID/VC patterns may yield interoperability advantages
Local LLM + hybrid routing is a primary battleground LM Studio support in NanoBot/PicoClaw, CoPaw's local judge routing, Moltis' context window overrides, vLLM/Ollama issues everywhere Self-hosted and edge deployments are no longer niche; agent frameworks must treat local model UX as first-class
Provider fragility is the #1 operational pain OpenClaw's Codex/OpenRouter regressions, Moltis' Gemini/Qwen patches, ZeroClaw's 5+ provider OAuth/errors Building robust provider abstraction layers with automatic failover, schema normalization, and auth resilience is a critical competitive moat
Memory reliability is a trust threshold NanoBot's dream overwrite bug, CoPaw's silent task stops, OpenClaw's OOM dreaming controls Users will abandon agents that lose or corrupt context; memory subsystems need data-loss guarantees, not just performance optimization
Enterprise IM integration = geographic market share LobsterAI's Chinese connector dominance, NanoBot's Feishu refinement, OpenClaw's global channel breadth Winning regional enterprise markets requires deep, native integration with local messaging platforms, not generic webhooks
Release quality is outpacing feature velocity as a success predictor OpenClaw's onboarding regression, IronClaw's staging P0s, ZeroClaw's S0 backlog The projects that survive the next 12 months will be those that slow feature expansion to stabilize onboarding and core reliability

Report compiled from 13 project digests dated 2026-04-16.

Peer Project Reports

NanoBot — HKUDS/nanobot

NanoBot Project Digest — 2026-04-16

1. Today's Overview

NanoBot saw very high development velocity in the last 24 hours, with 46 PRs updated (28 open, 18 merged/closed) and 17 issues active (11 open, 6 closed). No new release was cut. The project is in a fast-moving integration phase on the nightly branch: multiple large feature PRs are being split, reviewed, and re-landed (e.g., Microsoft Teams reintegration, performance optimizations from #3158). At the same time, a regression from v0.1.5 → v0.1.5.post1 was reported, indicating some stability risk around memory compatibility. Community engagement is strong, particularly from Feishu and Microsoft Teams users.

2. Releases

None today.

3. Project Progress

Merged/Closed PRs Today

PR Author Summary
#2600 T3chC0wb0y Microsoft Teams channel — landed on nightly after a long review cycle.
#2929 T3chC0wb0y Fixed Teams quote normalization hardcoded fallback and wired restart notifications.
#3178 LeoFYH Removed duplicate channel_id assignment in Discord message handler.
#3186 sohamb117 LM Studio first-class support + nullable API keys — closed after merge.
#3158 mohamed-elkholy95 Large omnibus performance/correctness PR closed and split into smaller PRs (#3180–#3184).
#3160 Aisht669 Added MiniMax Anthropic-compatible endpoint with reasoning_effort (thinking mode).
#2521 chengyongru MyTool runtime self-inspection tool — merged.

Key advances:

  • Channel expansion: Teams is now native; Feishu threading/reactions are being refined.
  • Local LLM UX: LM Studio is officially supported; null API keys are allowed.
  • Performance: A major optimization patchset was decomposed and is being re-reviewed for merge.
  • Agent introspection: Agents can now read/set their own runtime state via MyTool.

4. Community Hot Topics

Most Active Threads

# Item Comments Why It's Hot
#623 Persistent "Tool Not Found" Errors for Custom Skills 8 Skill development friction. New users building custom skills (Google Calendar, HA Voice) hit registration/discovery failures. Closed, but the pattern suggests docs or tooling gaps for skill authors.
#3123 Cron/scheduled task message send lacks user session 7 Core UX gap. Cron runs in an isolated session, so users cannot follow up on or refine automated outputs. Underlying need: persistent or attachable session context for scheduled agents.
#3143 LLM error 1214: illegal messages parameter 5 Provider reliability. High-frequency token-consolidation failures on Feishu. Suggests message-format validation or provider-specific adapter fragility.
#3107 一些建议 (Various suggestions) 5 Power-user configurability. Requests cover CLI model overrides, configurable timeouts, LLM fallback chains, and multi-provider setup. Signals that NanoBot is being adopted in team/enterprise settings where reliability and flexibility matter.

Notable Large PRs Under Review

  • #3144 — AgentHiFive integration (MCP-backed backend, vault-managed channels). This is a significant architectural addition and will likely draw sustained review attention.
  • #3191 — Teams reintegration onto latest nightly, incorporating all prior review feedback.

5. Bugs & Stability

Severity Issue Summary Fix PR?
🔴 Critical #3190 v0.1.5 → v0.1.5.post1 regression: crash on startup/memory load (Error processing message for session feishu:...). Reporter suspects memory-format incompatibility. None yet.
🟠 High #3143 Frequent LLM error 1214 during token consolidation on Feishu. Disrupts long conversations. None yet.
🟠 High #2957 MEMORY.md overwritten/blanked by "dream" — data loss bug. Closed today; fix presumably landed. Closed
🟡 Medium #3166 Feishu send_progress notifications missing. Fix likely in #3176. #3176
🟡 Medium #3102 NoneType vs int comparison in _build_kwargs when max_tokens/temperature are absent. Closed
🟡 Medium #3115 Cron deliver: false ignored when agent produces output. Closed
🟢 Low #2921 MSTeams restart notify config exposed but unimplemented. #3191

Stability assessment: The v0.1.5.post1 regression is the most urgent open risk. Memory/data-loss bugs (#2957) appear addressed, but the pattern raises confidence questions around the "dream" and consolidation subsystems.

6. Feature Requests & Roadmap Signals

Request Issue/PR Likelihood in Next Release
Per-phase model overrides (eval vs. exec in heartbeat) #3156 High — small, well-scoped, active review.
LM Studio native provider #3185 → #3186 Merged — already in.
MiniMax thinking mode #3160 Merged — already in.
Feishu thread-scoped sessions + reply_in_thread #3176 High — directly addresses enterprise Feishu UX.
MyTool runtime self-inspection #3177 High — already merged in #2521, this is a follow-up/reland.
AgentHiFive integration #3144 Medium — large PR, may need multiple review rounds.
LLM provider fallback + multi-custom provider #3107 Medium — architectural, but strongly requested by teams.
Configurable LLM timeout + CLI -model override #3107 Medium-High — small surface area, high user value.
WebSocket tooling + session lifecycle #3179 Medium — broad surface, may be staged.

7. User Feedback Summary

Pain Points

  • Memory/data-loss anxiety: The "dream" overwrite bug (#2957) and the v0.1.5.post1 crash (#3190) have users worried about MEMORY.md reliability.
  • Skill development is hard: Custom skill registration fails opaquely (📰 Hacker News AI Digest 2026-04-17 #623), creating a steep onboarding curve.
  • Cron/scheduled tasks feel disconnected: Users cannot iterate on cron outputs (#3123), limiting automation usefulness.
  • Provider robustness: Feishu users especially hit frequent LLM format errors (#3143) and lack progress feedback (#3166).
  • Billing blindness: No warning when an API key is in arrears (#3006).

Positive Signals

  • Strong enthusiasm from first-time personal-AI users (📰 Hacker News AI Digest 2026-04-17 #623: "very exciting getting things to work!").
  • Active enterprise/team adoption in China (Feishu) and globally (Teams, Telegram, Slack).
  • Appreciation for fast maintainer response: multiple issues closed within days of reporting.

8. Backlog Watch

These items have been open for weeks or longer and need maintainer triage or decision:

Item Age Why It Needs Attention
#2220 ~1 month Proposal: ContextVar for task-local tool routing. Async-safety hardening. Has architectural merit but no maintainer response.
#3006 ~6 days API-key arrearage silent failure. Small UX fix, zero engagement.
#3095 ~3 days Custom Anthropic API endpoint support. Blocked on provider architecture questions; no maintainer reply.
#3188 1 day nanobot-channel-webhook install fails from docs. Likely docs or packaging regression; needs quick verification.

Digest generated from HKUDS/nanobot GitHub activity for 2026-04-16.

Hermes Agent — nousresearch/hermes-agent

Hermes Agent Project Digest — 2026-04-16

1. Today's Overview

Hermes Agent saw very high activity on 2026-04-16 with 50 issues and 50 PRs updated in the last 24 hours, though no new releases were cut. The project is in an active stabilization and feature-polish phase: 28 issues were closed and 19 PRs merged/closed, with maintainers rapidly landing fixes for profile isolation, Docker compatibility, Slack/Discord gateway bugs, and provider edge cases. At the same time, 31 open PRs and 22 open/active issues indicate a healthy but backlogged pipeline. The community is particularly engaged around multi-agent architecture, security reporting, and AWS Bedrock support.

2. Releases

No new releases today.

3. Project Progress

Merged/Closed PRs Today (selected significant items)

PR Author Summary
#10570 teknium1 Fixed five HERMES_HOME profile-isolation leaks — honcho session titles, hardcoded skill paths, install.sh --dir, honcho config fallback, and skill manager external dirs write-through.
#10569 teknium1 Fixed Docker ps compatibility — replaced BSD ps -ax with POSIX ps -A to prevent false "gateway not running" reports in containers.
#6644 shin4 Passed HERMES_HOME to execute_code subprocess — fixes auth path resolution for tools like vision_analyze in Docker.
#6527 iacker Made install.sh --dir honor HERMES_HOME state — seeds correct profile-scoped skill directory during install.
#6338 iacker Blocked mutations in external skill dirs — prevents edit/patch/delete flows from touching skills.external_dirs.
#6025 dontcallmejames Fixed honcho config fallback respecting HERMES_HOME — stopped cross-profile memory leakage.

Features Advanced

  • Rich terminal output: The stacked PRs for intra-line diff view, syntax highlighting, and markdown/theming support (Issue #4518) were closed, indicating the layered terminal rendering engine in agent/rich_output.py has landed.
  • Web UI Gateway: The umbrella feature for a local browser-based interface (Issue #501) was closed, suggesting substantial progress or completion.
  • Multi-Agent Architecture: The foundational umbrella issue (Issue #344) was closed after 27 comments — a major architectural milestone.

4. Community Hot Topics

Item Status Comments 👍 Analysis
#344 — Multi-Agent Architecture CLOSED 27 13 The most discussed issue in the dataset. Signals strong community demand for orchestration, specialized roles, and resilient workflows beyond single-agent delegation. Closing this umbrella issue suggests either a roadmap shift or incremental delivery via sub-issues.
#9179 — SECURITY.md / private vulnerability reporting OPEN 24 0 Unusually high engagement for a process issue. Reflects growing production adoption and security researcher interest. The project currently lacks GitHub private vulnerability reporting, which is becoming a baseline expectation.
#3863 — Native AWS Bedrock provider support CLOSED 13 17 High-vote enterprise feature request. Closed status may indicate resolution via OpenRouter or a provider plugin; the cost/latency/VPC concerns remain relevant for AWS-heavy users.
#501 — Web UI Gateway CLOSED 13 1 Closed despite low votes, likely because core implementation merged. Competitor parity (Claude Artifacts, OpenAI Canvas) is a clear driver.
#4518 — Intra-line Diff-View, Syntax Highlighting, Theming CLOSED 13 0 Developer-experience polish that closed via 5 stacked PRs. Shows investment in terminal UI quality as a differentiator.

5. Bugs & Stability

Severity Issue Status Fix PR? Description
High #9179 OPEN ❌ No private vulnerability reporting channel — security process gap.
High #10223 OPEN ❌ GitHub Copilot provider returns empty base URL via credential pool — breaks startup entirely.
Medium #10575 OPEN ✅ #10576 Anthropic OAuth/Claude Max proxy falsely reports "extra usage exhausted" due to system prompt misclassification. Fix PR open by same reporter.
Medium #6843 CLOSED ✅ UnicodeEncodeError on API calls with zai provider — fixed.
Medium #7237 CLOSED ✅ "Response truncated due to output length limit" breaking long-form CLI/gateway responses — fixed.
Medium #7588 CLOSED ✅ Complete tool failure (memory, file read, history lookup) — fixed.
Medium #6360 CLOSED ✅ Quick install custom model fails with Invalid port: 6153export — setup script bug, fixed.
Medium #6447 OPEN ❌ Discord /sethome writes channel ID to config.yaml instead of .env, breaking env-based config workflows.
Medium #10581 OPEN ❌ Home-channel auto-prompt only checks env var, never yaml config — compounds #6447.
Low #10579 CLOSED ✅ Slack mention prefix breaks is_command() slash-command detection — fixed.
Low #10550 CLOSED ✅ Discord /skill group exceeds 8000-byte limit, causing fatal CommandSyncFailure on startup — fixed.
Low #10454 CLOSED ✅ Feishu (Lark) occasionally drops last message in conversation — fixed.

6. Feature Requests & Roadmap Signals

Issue Votes Prediction
#3863 — Native AWS Bedrock provider 17 Likely near-term. Enterprise demand is strong; closed status may mask ongoing provider work.
#10299 — Cascading context file discovery 0 Moderate priority. Aligns with multi-repo and monorepo developer workflows; low friction to implement.
#10488 — Secure first-run web onboarding wizard 0 Medium-term. Complements the closed Web UI Gateway issue; broadens non-technical user adoption.
#10585 — Multi-resolution context compression (MEMORY.md / USER.md) 0 High impact if implemented. Cost reduction is a persistent theme; may be explored as token prices and context windows evolve.
#10583 — Route Slack reaction_added events to agent 0 Small, likely soon. Tight Slack integration fix; fits current gateway polish trajectory.
#10513 — WhatsApp default to self-chat instead of bot 0 Easy UX win. One-line default change; strong user-experience justification.

7. User Feedback Summary

Pain Points

  • Setup fragility: Multiple closed issues around quick install (#6360), HERMES_HOME isolation leaks (#5947), and Docker ps behavior (#9723) show users struggle with non-default environments.
  • Gateway platform bugs: Slack (#10579, #10583, #10582), Discord (#6447, #10550), and Feishu (#10454) all had active bugs — messaging platform reliability is a critical path.
  • Provider edge cases: Copilot empty base URL (#10223), Anthropic OAuth false exhaustion (#10575), and Unicode errors (#6843) indicate provider abstraction layer still has brittle spots.
  • Context/cost bloat: Issue #10585 explicitly calls out MEMORY.md / USER.md loading fully into every call as a "massive cost driver."

Positive Signals

  • Strong engagement on architectural roadmap items (multi-agent, Web UI, rich terminal output).
  • Rapid maintainer response time: many issues created in early April were closed within days.
  • Active community contributing stacked PRs for complex features (e.g., #4518).

8. Backlog Watch

Item Age Risk Notes
#9179 — SECURITY.md / private vulnerability reporting ~3 days High 24 comments, no maintainer resolution. Security process gaps become blockers for enterprise adoption.
#10223 — GitHub Copilot provider empty base URL ~1 day High 4 upvotes, startup-breaking. No linked fix PR yet.
#6447 — Discord /sethome config leak ~7 days Medium Known bug with clear reproduction; may need gateway config refactor.
#10160 — Nous OAuth cross-process sync ~1 day Medium Open PR preventing cron job session revocation. Needs review.
#4692 — Ink/React TUI refactor ~13 days Medium Large new terminal UI PR. High impact but needs careful review.
#4424 — Honor main tracking remote during update checks ~15 days Low Contributor workflow improvement; stalled without maintainer feedback.
PicoClaw — sipeed/picoclaw

PicoClaw Project Digest — 2026-04-16

1. Today's Overview

PicoClaw shows very high development velocity with 26 PRs updated in the last 24 hours (20 merged/closed, 6 open) and 7 active issues. The project released a new nightly build (v0.2.6-nightly.20260415.c0fadc59), indicating steady progress toward the next patch release. Activity spans core agent behavior, channel integrations, web UI polish, and cross-platform stability fixes. The maintainer team appears highly responsive, with same-day turnaround on several PRs. Overall project health looks strong and accelerating.

2. Releases

  • v0.2.6-nightly.20260415.c0fadc59 — Nightly Build
    • Automated nightly ahead of v0.2.6; no explicit changelog beyond main delta.
    • Caution: Marked unstable; typical for automated builds.
    • No breaking changes or migration notes published.

3. Project Progress

Merged/Closed PRs Today (20 total, selected highlights)

PR Author Summary
#2535 afjcjsbx MCP slash commands (/list mcp, /show mcp) for chat channels
#2531 is-Xiaoen delegate tool for cross-agent task handoff (Phase 2, #2148)
#2532 lxowalle /btw side-question command — immediate ask without touching session history
#2503 cytown Parallel agent loop refactor with updated docs
#2413 ex-takashima LINE channel migrated to official SDK v8
#2524 SiYue-ZO Sogou-backed web search for mainland China users + web UI config panel
#2528 wj-xiao Fix: preserve reused tool call IDs across turns for strict providers
#2529 lc6464 Markdown syntax highlighting in web chat and skills preview
#2530 wj-xiao Fix: nested channel_list patches for channel config saves
#2502 lxowalle /btw one-off side-question command (preceding #2532)
#2525 afjcjsbx Fix: recover after image-input-unsupported model failures
#2526 wj-xiao Fix: restore chat composer disabled-state messaging
#2403, #2422 wj-xiao Gateway PID liveness hardening and stale pidfile cleanup
#2417 wj-xiao Fix: disable seahorse context manager on freebsd/arm
#2434 wj-xiao Fix: skip isolation env test on unsupported OSes
#2466 wj-xiao Fix: fallback to token auth on unsupported platforms
#2467 wj-xiao Fix: align react/react-dom versions in launcher
#2484 BeaconCat LLM-as-Judge evaluation mode for membench
#2500 cytown Remove useless backend output for platform-token

Key advances: Multi-agent delegation, MCP server UX, parallel execution, China-localized search, and robust cross-platform stability.

4. Community Hot Topics

Item Activity Underlying Need
#28 Feat Request: LM Studio Easy Connect 12 comments, 👍 1 Local/self-hosted LLM accessibility. Users want frictionless connection to LM Studio, especially on Android/edge devices. Signals demand for broader local-provider onboarding.
#2468 Scheduled Task Fails to Execute 5 comments Cron tool permission model clarity. Error "scheduling command execution is restricted to internal channels" suggests users hit policy boundaries without clear guidance. Needs better docs or channel-scope configuration.
#2046 PicoClaw does not call tool with LongCat API 2 comments Provider-specific tool-calling compatibility. LongCat API integration is broken; points to incomplete provider adapter coverage.
#2302 Web UI requires manual re-authentication frequently 1 comment Credential/session persistence reliability. Antigravity API PERMISSION_DENIED erodes trust in unattended setups.
#2446 Messages echoed back in multi-channel setup 1 comment, 👍 1 Cross-channel state isolation. Pending tasks in one channel leak into another's response path — a concurrency/routing bug.

5. Bugs & Stability

Severity Issue/PR Description Fix PR?
High #2468 Cron tool execution blocked by internal-channel restriction None linked
High #2446 Multi-channel message echo / cross-channel task contamination None linked
Medium #2302 Frequent credential re-auth in Web UI (antigravity API) None linked
Medium #2046 Tool calls fail with LongCat API provider None linked
Medium #2528 (merged) Reused tool call IDs dropped globally → provider sequencing failures Fixed
Medium #2525 (merged) Session stuck after sending images to non-vision models Fixed
Low #2530 (open) Channel config saves missing nested channel_list patches In review
Low #2526 (merged) Chat composer disabled-state messaging broken Fixed

Note: Four open bugs lack linked fix PRs and represent the day's largest stability gaps.

6. Feature Requests & Roadmap Signals

Request Issue/PR Likelihood in Next Version
LM Studio easy connect / local provider UX #28 Medium — long-standing request, high comment volume, but requires dedicated provider work
--no-browser flag for picoclaw auth login #2533 High — small CLI flag, clear use case (headless/automated setups)
Configurable fresh_tail_size in Seahorse #2527 High — one-constant change, already scoped to config system
MCP slash commands in chat #2535 Very High — PR already open, aligns with MCP momentum
Cross-agent delegate tool #2531 Very High — core Phase 2 roadmap item, PR open
Parallel agent loop #2503 Very High — architectural refactor with docs, builds on merged #2481

7. User Feedback Summary

Pain Points

  • Provider fragility: LongCat API tool-calling broken; antigravity API sessions expire unpredictably.
  • Authentication friction: Browser-pop auth flow unsuitable for containers/headless deploys; frequent Web UI re-auth.
  • Multi-channel reliability: Task state bleeds across channels, breaking isolation expectations.
  • Local/self-hosted LLM gap: LM Studio integration too difficult for non-technical users.

Positive Signals

  • Responsiveness: Same-day merges on multiple PRs.
  • China UX attention: Sogou search integration shows geographic user-base awareness.
  • Advanced features landing: MCP commands, agent delegation, and parallel loops signal mature architecture.

8. Backlog Watch

Item Age Risk Notes
#28 LM Studio Easy Connect ~2 months High Highest-comment open issue; no linked PR. Represents accessibility barrier for local-LLM users.
#2046 LongCat API tool-calling bug ~3 weeks Medium Provider compatibility gap; only 2 comments, may be under-prioritized.
#2302 Antigravity API re-auth loop ~2 weeks Medium Impacts unattended deployments; no fix PR visible.
#2413 LINE SDK v8 refactor ~1 week Low Open but updated today; likely nearing merge.

Maintainer attention recommended: #28 (local LLM onboarding) and #2468/#2446 (stability bugs with no fixes).

NanoClaw — qwibitai/nanoclaw

NanoClaw Project Digest — 2026-04-16

1. Today's Overview

NanoClaw saw high development velocity in the last 24 hours with 17 PRs updated (13 open, 4 merged/closed) and 5 fresh issues opened, all currently active. The project is in a feature-heavy expansion phase, with significant work on LLM provider diversification (OpenAI, OpenCode, multi-provider via Vercel AI SDK), container skill additions, and infrastructure hardening (scheduler fixes, channel isolation, pnpm migration, Node 24 upgrade). No new releases were cut. Notably, all 5 new issues come from a single reporter (ythx-101), suggesting concentrated onboarding friction around /setup, mount semantics, and container path handling.

2. Releases

No new releases.

3. Project Progress

Merged/Closed PRs (4 items)

PR Author Summary
#1281 heyjawrsh Headless Linux browser fallback + VPS docs: openBrowser() now prints URL to stdout instead of silently failing on $DISPLAY-less Linux servers.
#1777 johnnyfish Forward ONECLI_API_KEY to SDK: Fixes authenticated container config for cloud/dev OneCLI gateways.
#1782 openclaw-shi Credential proxy ETIMEDOUT fix: TCP keepalive, cleaner connection teardown, and upstream error handling for streaming.
#1760 Viral1010 Twilio fallback calls: Voice reminders now retry a fallback number if the primary call fails.

Key advances: Deployment reliability for headless/VPS environments improved; voice call robustness increased; authentication and streaming proxy stability patched.

4. Community Hot Topics

No items have comments or reactions yet, so "hot" is determined by strategic significance and author activity:

Item Link Why It Matters
PR #1786 — 7-layer scope expansion roadmap qwibitai/nanoclaw#1786 Massive feature dump (14 items) across webhook sources, LLM validation, confidence calibration, and cross-reference verification. Signals aggressive roadmap execution.
PR #1784 — Multi-provider LLM layer via Vercel AI SDK qwibitai/nanoclaw#1784 Core architecture change enabling OpenAI, Google, Ollama, Groq, Together. Dual-runtime design preserves Anthropic path.
PR #1771 — pnpm migration qwibitai/nanoclaw#1771 Infrastructure change touching CI, Dockerfile, setup scripts. High blast radius if merged.

Underlying needs: The community and core contributors are pushing for vendor independence (multiple LLM providers), operational maturity (better package management, container upgrades), and broader automation layers (webhooks, pre-meeting briefings, validation pipelines).

5. Bugs & Stability

Severity Issue/PR Description Fix PR?
High #1788 Scheduler silent hot-loop from runTask throws and non-text scripts. Three-layer defense: write-time coercion, startup repair, runtime guard. Yes — PR open
High #1785 Single broken channel (e.g., expired Gmail OAuth) crashes entire service startup. Yes — PR open
Medium #1787 /setup on macOS with Apple Container produces 6 git merge conflicts on first run. No
Medium #1789 Non-ASCII display names silently normalized to dm-with-unnamed. No
Medium #1790 /setup mount allowlist semantics unclear (directory vs. file). No
Medium #1791 Mount allowlist can't differentiate peer files in same directory across groups. Security gap. No
Low-Medium #1792 additionalMounts.containerPath must be relative; absolute paths rejected silently/undocumented. No

Stability assessment: Two critical runtime fixes are in open PRs. Onboarding (/setup) and mount security have emerging but unpatched issues.

6. Feature Requests & Roadmap Signals

Signal Source Likely Next-Version Inclusion?
Multi-provider LLM support PRs #1784, #1774, #1776 Very high — 3 parallel PRs indicate this is a top priority.
OpenCode agent provider PR #1776 High — aligned with v2 provider model.
Composio MCP for managed OAuth PR #1781 High — reduces Gmail/Calendar setup friction.
5 new marketing/container skills PR #1780 Moderate — skill PRs typically merge faster than core infra.
News briefing skill PR #886 Moderate — open since March 9, needs review.
pnpm + Node 24 upgrade PRs #1771, #1778 High — infra debt, blocking some dependencies.

7. User Feedback Summary

Real pain points from ythx-101's issue cluster:

Pain Point Issue Implication
Opaque mount semantics #1790, #1792 Users need trial-and-error to configure container mounts. Documentation and validation messages are inadequate.
Security granularity gap #1791 Cannot share one file in a directory while hiding another. Multi-tenant/token scenarios are under-supported.
Localization/Unicode handling #1789 Non-ASCII names silently degraded. Suggests ASCII-only assumptions in folder naming logic.
macOS onboarding broken #1787 Apple Container branch merge-conflicts on first setup. Major friction for macOS developers.

Satisfaction/dissatisfaction: The reporter is actively testing edge cases and documenting failures clearly — a sign of engaged but frustrated early adoption. The lack of maintainer responses (0 comments) on all 5 issues is a community health risk.

8. Backlog Watch

Item Age Risk Action Needed
PR #886 — Daily news briefing skill ~38 days (Mar 9) Stalled feature contribution with clear value Maintainer review/merge decision
Issues #1787–#1792 (all from ythx-101) 1 day Rapid cluster of onboarding bugs with zero maintainer engagement Triage and at least acknowledge; #1787 and #1789 are likely quick fixes
PR #1311 — Create new session ~26 days (Mar 21) Feature skill PR idle Review or request changes

Priority recommendation: Address the 5-issue cluster from ythx-101 first — it represents a concentrated onboarding failure mode that could deter new contributors.

NullClaw — nullclaw/nullclaw

NullClaw Project Digest — 2026-04-16

1. Today's Overview

NullClaw saw high community activity in the last 24 hours with 9 active issues and 8 pull requests updated, though no new releases were cut. The project is experiencing growing pains around build tooling (Zig version migrations, Debian installation friction), runtime stability (shell execution on Linux, HTTPS constraints breaking OTel), and polish gaps (version reporting, CLI health check inconsistencies). Three PRs were merged or closed, indicating maintainers are actively landing reviewed work, but the open PR backlog remains substantial with several ready-for-review contributions.

2. Releases

No new releases.

3. Project Progress

Merged/Closed PRs (3 items)

PR Author Summary
#369 jmylchreest Signal "Note to Self" support — adds parseSyncNoteToSelf() to handle sync envelopes instead of blanket-dropping them, with 4 new tests.
#377 jmylchreest Sandbox-validated PATH environment variables — introduces tools.path_env_vars config, passing colon-separated paths to shell children after sandbox validation.
#378 jmylchreest Security fix for git -C + Signal Note to Self — tightens allowlist handling so uppercase -C is not misclassified as dangerous lowercase -c; also bundles the Signal sync fix.

What advanced: Signal integration matured with self-messaging support, and shell sandboxing gained finer-grained environm

⚠️ 内容超过 GitHub Issue 上限，完整报告见提交的 Markdown 文件。

Metadata

Metadata

Assignees

No one assigned

    Labels

    openclaw-en

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions