Skip to content

🦞 OpenClaw Ecosystem Digest 2026-03-26 #294

@github-actions

Description

@github-actions

OpenClaw Ecosystem Digest 2026-03-26

Issues: 500 | PRs: 500 | Projects covered: 13 | Generated: 2026-03-26 00:11 UTC


OpenClaw Deep Dive

OpenClaw Project Digest — March 26, 2026

1. Today's Overview

OpenClaw shows exceptionally high activity with 500 issues and 500 PRs updated in the last 24 hours—indicating a rapidly evolving codebase with significant community engagement. The project released three versions (v2026.3.24 and two betas) focusing on OpenAI API compatibility and security hardening. However, regression risk is elevated: numerous reports cite breakages after the v2026.3.22 upgrade, particularly around OAuth token handling, Discord multi-account startup, and session management. The maintainer team appears responsive with same-day PR merges, but the velocity of changes is outpacing some users' ability to adapt. Cross-platform expansion (Linux GTK4 app) and enterprise features (Model Router, episodic memory) signal maturation toward production deployments.


2. Releases

v2026.3.24 (Stable)

Aspect Details
Breaking Changes None explicitly flagged
Key Changes • OpenAI-compatible gateway endpoints: /v1/models, /v1/embeddings, and explicit model override forwarding through /v1/chat/completions and /v1/responses (broader RAG/client compatibility) — #vincentkoc
• Agents/tools: /tools endpoint now exposes available tools
Migration Notes Users relying on implicit model routing should verify explicit overrides work as expected

v2026.3.24-beta.2

Aspect Details
Fixes • Outbound media/local files: aligned with fs policy—host-local files and inbound-media paths send when workspaceOnly is off; strict workspace-only agents remain sandboxed
• Runtime/install: lowered Node 22 floor to 22.14

v2026.3.24-beta.1

Aspect Details
Changes Same OpenAI compatibility additions as stable (beta precursor)

3. Project Progress

Merged/Closed PRs Today (Selected)

PR Author Summary Impact
#54453 ikebo Make message tool buttons schema property optional Fixes Telegram plugin schema validation failures
#54679 gumclaw Allow MS Teams feedback/welcome config keys Unblocks Teams channel configuration
#54694 drobison00 Block silent reconnect scope-upgrade escalation Security hardening for gateway auth
#54418 adzendo Make buttons schema optional in message tool (duplicate fix) Redundant safety for Telegram
#53401 erhhung Fix tools.web.fetch.maxResponseBytes schema validation Resolves config parsing regression
#54575 w-sss Add buildToolContext for Telegram forum topic thread IDs Fixes exec approval routing in forum topics
#45527 MastrXplorer Resolve symlinks before skill directory scanning Security fix for path traversal via symlinks

Active Development (Open PRs)

  • Linux GTK4 Companion App #53905 — Major platform expansion
  • Model Router #54562 — Automatic tier-based model selection
  • Episodic Memory System #54750 — CLS-inspired hippocampal memory for agents
  • SambaNova Provider #54748 — New inference provider integration

4. Community Hot Topics

Top Issues by Engagement

Issue Comments 👍 Core Need
#75 Linux/Windows Clawdbot Apps 49 65 Cross-platform parity — Users want native desktop apps matching macOS feature set; Linux PR #53905 in progress
#52885 WeChat plugin incompatible with 2026.3.22+ 40 0 Plugin ecosystem stability — Third-party plugins breaking on core updates; needs stable plugin SDK contract
#52823 Control UI assets not found (closed) 34 12 Developer experience — Build step documentation gaps; fix merged
#44851 Kimi web_search 401 auth error 29 3 Multi-provider auth consistency — Tool-level auth differs from chat API
#14593 Skill install fails in Docker: brew not installed 16 12 Container-ready deployment — Brew-based skills incompatible with Linux containers

Underlying Needs Analysis

  • Platform expansion urgency: 65 upvotes on Linux/Windows apps shows this is the 📋 Claude Code 社区日报 2026-02-23 #1 user-facing gap
  • Plugin API stability: WeChat breakage exemplifies tension between rapid core evolution and third-party ecosystem
  • Auth fragmentation: Multiple OAuth/token refresh issues suggest need for unified credential management

5. Bugs & Stability

Severity Issue Status Fix PR
🔴 Critical #53317 Gateway overwrites fresh OAuth token with stale cache on startup (v2026.3.22, openai-codex) Open None identified
🔴 Critical #52037 OpenAI Codex OAuth token refresh not persisting to disk Open None identified
🟡 High #53132 Discord multi-account gateway hangs at 'awaiting gateway readiness' (v2026.3.22) Open None identified
🟡 High #53644 Discord WebSocket 1006 crashes gateway instead of reconnecting Closed #53644
🟡 High #52875 Session_send gives "no session found" after 2026.3.22 upgrade Open None identified
🟡 High #53870 Matrix plugin blocked by pending VirusTotal security scan Open None identified
🟢 Medium #53244 WhatsApp plugin runtime modules missing from npm package (2026.3.22+) Open None identified
🟢 Medium #54514 npm package missing docs/reference/templates Open None identified

Regression Pattern

v2026.3.22 appears to be a problematic release: Multiple independent reports of session, OAuth, and Discord breakages starting with this version. Recommend hotfix release or rollback guidance.


6. Feature Requests & Roadmap Signals

Feature Issue/PR Likelihood in Next Version Rationale
Native MCP Client support #29053 High 10 upvotes, aligns with industry standardization (Anthropic MCP); would reduce tool fragmentation
Linux native app #75 + #53905 In Progress PR open, v1 implementation complete
Model Router (automatic tier selection) #54562 High PR open, addresses cost optimization demand
Episodic memory system #54750 Medium Large PR, novel architecture—may need extended review
Async exec callback #18237 Medium 5 upvotes, subagent workflow pain point
Agent identity/trust verification (ERC-8004/DID) #49971 Low Speculative, standards-aligned but niche demand

7. User Feedback Summary

Pain Points

Theme Evidence Severity
Upgrade fragility Multiple "worked before 2026.3.22" regressions High
OAuth/token management #53317, #52037, #26322, #36982 High
Documentation/build gaps #52823, #54514 Medium
Docker/container deployment #14593 brew dependency Medium
Token waste in prompts #9157 93.5% budget waste Medium

Positive Signals

  • Strong engagement: 500 issues/PRs in 24h shows healthy community
  • Responsive maintainers: Same-day merges on critical fixes
  • Enterprise readiness: Model Router, memory systems, multi-account Discord show scale ambitions

8. Backlog Watch

Issue Age Issue Why It Needs Attention
#75 ~3 months Linux/Windows apps 65 upvotes, PR #53905 ready—needs final review/merge
#9157 ~2 months Token budget waste 12 upvotes, performance-critical, no PR
#18237 ~1.5 months Async exec callback Subagent workflow blocker, no PR
#29053 ~1 month MCP Client support Industry alignment opportunity, no PR
#10008 ~2 months Feishu plugin re-registration Performance regression, no fix identified

Maintainer Action Recommended

  1. Hotfix for v2026.3.22 regressions — OAuth and session issues are degrading trust
  2. Finalize Linux app PR — High community demand, implementation complete
  3. Plugin SDK stability commitment — Document and communicate breaking change policy to third-party developers

Cross-Ecosystem Comparison

Cross-Project AI Agent Ecosystem Report — March 26, 2026

1. Ecosystem Overview

The open-source personal AI assistant landscape is experiencing intense consolidation velocity, with top projects processing 50–500+ GitHub events daily as they race to stabilize core agent loops while expanding modality support. The field has bifurcated into "batteries-included" platforms (OpenClaw, LobsterAI, CoPaw) competing on channel breadth and enterprise features versus minimalist/specialized frameworks (NanoBot, ZeptoClaw, TinyClaw) optimizing for specific deployment contexts. A critical inflection point is emerging around MCP (Model Context Protocol) adoption, multi-tenant isolation, and local/edge inference—capabilities that are becoming table stakes for production deployment. Supply chain security (NanoBot's LiteLLM quarantine) and platform parity (Linux/Windows native apps) are now dominant community concerns, reflecting maturation from hobbyist to enterprise-adjacent use cases.


2. Activity Comparison

Project Issues (24h) PRs (24h) Release Status Health Score Notes
OpenClaw 500 500 v2026.3.24 (stable) + 2 betas B+ Exceptional velocity; regression risk from v2026.3.22
NanoBot 18 51 None (accumulating) B Post-LiteLLM stabilization sprint
ZeroClaw 39 50 v0.6.2 (prior) B+ Pre-patch stabilization for v0.6.3
PicoClaw 182 277 v0.2.4 (2026-03-25) A- 93% issue closure rate; stabilization phase
NanoClaw 37 50 None B CLAUDE.md fixes merged; Linux fragility emerging
NullClaw 3 3 None B Moderate velocity; WeChat/vision demand
IronClaw 10 50 v0.22.0 imminent B+ v2 engine + multi-tenant isolation in flight
LobsterAI 37 50 2026.3.25 (latest) B Critical crash loops; MCP reliability issues
TinyClaw 0 3 None B+ Infrastructure modernization; low issue volume
Moltis 3 3 None B+ Moderate velocity; local thinking demand
CoPaw 50 49 v0.2.0.post1 B v0.2.0 upgrade fragility; China market focus
ZeptoClaw 11 30 None A- Rapid integration; all critical bugs closed
EasyClaw 0 0 No activity

3. OpenClaw's Position

Advantages vs. Peers

Dimension OpenClaw Position Peer Gap
Scale & ecosystem 10× activity volume of nearest competitor; 500 issues/PRs daily NanoBot, ZeptoClaw operate at 1/10th scale
Enterprise features Model Router, episodic memory, multi-account Discord IronClaw catching up on multi-tenant; others lack
Cross-platform Linux GTK4 app in progress (#75: 65 upvotes) PicoClaw, NanoClaw have Ollama/local focus but weaker channel breadth
OpenAI compatibility Native gateway endpoints (/v1/*) LobsterAI has compatibility gaps; NanoBot rebuilding post-LiteLLM

Technical Approach Differences

  • vs. NanoBot: OpenClaw maintains abstraction layer stability; NanoBot's LiteLLM removal forced architectural rupture
  • vs. IronClaw: OpenClaw prioritizes feature velocity; IronClaw invests in formal verification (multi-tenant compile-time TenantCtx)
  • vs. ZeptoClaw: OpenClaw is platform-agnostic; ZeptoClaw optimizes for Telegram-first deployment with deep UX integration

Community Size Comparison

OpenClaw's 500 daily events dwarfs all peers. Second tier: PicoClaw (459), CoPaw (99), IronClaw (60). OpenClaw's challenge is managing scale—v2026.3.22 regressions show velocity outpacing quality assurance.


4. Shared Technical Focus Areas

Requirement Projects Specific Needs
MCP (Model Context Protocol) tooling LobsterAI (#728, #820), NanoBot (#29053), Moltis (#439) "0 tools" discovery failures; dev/prod parity; custom display names
Multi-tenant isolation IronClaw (#1614), OpenClaw (enterprise features), CoPaw (#2283) Compile-time tenant contexts; per-agent model configs; session isolation
Local/edge inference PicoClaw (#75 Ollama), NanoBot (Step Fun), Moltis (#489/#490 local thinking), NullClaw (#624 vision) Cost reduction; privacy; 120s+ timeout handling; model diversity
Cross-platform native apps OpenClaw (#75 Linux/Windows), NanoClaw (#1075 Linux docs), TinyClaw (#260 shell→Node.js) GTK4, macOS menu bar, container-optimized builds
OAuth/token management OpenClaw (#53317, #52037), NanoBot (#2465), IronClaw Unified credential lifecycle; refresh persistence; scope escalation prevention
Memory/scalability architecture CoPaw (#2289 MEMORY.md), PicoClaw (episodic memory #54750), IronClaw (full transcript persistence) Vector DB migration; bounded history; context overflow detection

5. Differentiation Analysis

Project Primary Differentiator Target User Architecture Signature
OpenClaw Feature velocity + channel breadth Power users, enterprises Modular skills, gateway abstraction, rapid iteration
NanoBot Supply-chain resilience, multi-language Security-conscious, international Post-LiteLLM native providers, Chinese community
ZeroClaw A2A protocol, GitHub integration Multi-agent orchestrators Agent-to-agent interoperability focus
PicoClaw Voice/ASR-TTS, subagent workflows Conversational AI, automation SubTurn execution, hot-reload config
IronClaw Formal multi-tenant isolation, NEAR AI Web3, regulated environments Compile-time tenant contexts, WASM sandboxing
LobsterAI Commercial monetization, enterprise IM SaaS operators, China market Credit packs, DingTalk/POPO integration
ZeptoClaw Telegram-native UX, browser automation Community bots, rapid deployment agent-browser integration, threading optimization
TinyClaw Minimalist Node.js, office UI Individual developers, teams Pure JS, reduced dependencies, workspace metaphor
CoPaw Multi-agent collaboration, China channels Enterprise automation, stock monitoring Async agent communication, Xiaoyi/Feishu focus

6. Community Momentum & Maturity

Rapid Iteration Tier (>100 events/day, high feature velocity)

Project Trajectory Risk Profile
OpenClaw Aggressive expansion Regression debt from v2026.3.22
PicoClaw Feature→stabilization transition Exec tool rollback shows caution
CoPaw Post-release stabilization v0.2.0 upgrade path fragile

Stabilization Tier (50–100 events/day, quality focus)

Project Trajectory Key Investment
IronClaw v2 engine + security hardening Multi-tenant isolation, anti-drift monitoring
NanoBot Post-architectural recovery Native provider ecosystem, sandboxed execution
ZeptoClaw Bug closure excellence Browser automation GA

Maintenance/Modernization Tier (<50 events/day)

Project Trajectory Observation
LobsterAI Critical issue backlog Crash loops (#859) need immediate triage
TinyClaw Infrastructure refresh Low issue volume suggests under-reporting or stability
Moltis Community contribution mode openai-oxide PR signals maintainer burden

Dormant

Project Status
EasyClaw No activity; likely abandoned or private development

7. Trend Signals for AI Agent Developers

1. MCP as Emergent Standard

Multiple projects (LobsterAI, NanoBot, Moltis) investing in MCP client/server support. Signal: Tool fragmentation consolidating around Anthropic's protocol; custom tool ecosystems becoming technical debt.

2. Supply Chain Security as First-Class Concern

NanoBot's LiteLLM quarantine triggered urgent native provider migration. Signal: Dependency minimization and audit transparency becoming selection criteria; "vendored" providers preferred over abstraction layers.

3. Voice/Multimodal as Next Battleground

PicoClaw's ASR/TTS refactor (#1939), NullClaw's vision pipeline request (#624), LobsterAI's desktop features. Signal: Text-only agents commoditizing; voice + vision differentiation emerging.

4. Platform Parity as Growth Ceiling

OpenClaw's #75 (65 upvotes), NanoClaw's Linux documentation mismatch (#1075), TinyClaw's shell→Node.js migration. Signal: macOS-first development creates friction; container-native and cross-platform tooling winning.

5. Memory Architecture as Scaling Bottleneck

CoPaw's MEMORY.md scalability (#2289), PicoClaw's episodic memory, IronClaw's transcript persistence. Signal: File-based memory insufficient; vector DB + structured storage migration required for production load.

6. Enterprise IM as Regional Moat

LobsterAI (DingTalk/POPO), CoPaw (Xiaoyi/Feishu), OpenClaw (WeChat). Signal: Slack/Discord insufficient for global markets; China-specific integrations becoming competitive requirement.

7. Agent Orchestration vs. Single-Agent Focus

ZeroClaw's A2A protocol, CoPaw's multi-agent async, IronClaw's routine engine. Signal: Single-agent tools maturing; multi-agent coordination the next capability frontier.


Report generated from 12 project digests covering 1,500+ GitHub events on 2026-03-26.


Peer Project Reports

NanoBotHKUDS/nanobot

NanoBot Project Digest — 2026-03-26

1. Today's Overview

NanoBot shows exceptionally high development velocity with 51 PRs and 18 issues updated in the last 24 hours, indicating an active, rapidly evolving codebase. The project is undergoing significant architectural transitions—most notably the removal of LiteLLM due to supply chain security concerns—which is driving a wave of native provider implementations and compatibility fixes. No new releases were cut today, suggesting the maintainers are accumulating changes before a version bump. The community is actively engaged with multi-language participation (English, Chinese), and there's clear tension between rapid feature expansion and stability needs, particularly around provider compatibility and message handling reliability.


2. Releases

No new releases (0 today). Latest release remains prior to this period.


3. Project Progress

Merged/Closed PRs Today

PR Description Impact
#2472 Add Step Fun (阶跃星辰) provider support New native provider; responds to #2470
#2478 Message send retry mechanism with exponential backoff Reliability improvement for all channels
#2477 Configurable timezone for runtime context Precision for scheduling/time-aware agents
#1680 Mistral Provider + general transcription service Closed after transcription improvements merged
#1610 Auto-cast parameter types to match tool schema Robustness for tool calling

Key Advances:

  • Provider ecosystem expansion: Step Fun added; AWS Bedrock (#2485) and Discord migration (#2486) in progress
  • Infrastructure reliability: Retry logic, timezone precision, parameter validation
  • Post-LiteLLM stabilization: Multiple PRs addressing the architectural transition

4. Community Hot Topics

Rank Item Comments Analysis
1 #2208 Telegram message duplication on Android 8 Platform-specific messaging bug—Android Telegram client behavior differs from Desktop/Web; suggests channel abstraction layer needs platform-aware deduplication
2 #2456 nanobot onboard system deadlock in multi-user Linux 7 Critical infrastructure risk—SSH lockout in shared environments indicates unsafe system-level operations; needs sandboxing review
3 #2443 LiteLLM quarantined (credential stealer) 6 Supply chain security crisis—drives urgent provider migration; community needs migration guide
4 #2463 Prompt prefix not preserved in conversation history 6 Architectural correctness—breaks OpenAI prompt caching assumptions; affects reproducibility and cost optimization
5 #2469 Qwen-3.5 function.arguments JSON format error 5 Model-specific compatibility—code model parameter formatting issue

Underlying Needs:

  • Trust & security: LiteLLM incident has shaken confidence; users need transparency on dependency audits
  • Reliable multi-platform messaging: Telegram issues reveal gaps in cross-platform message handling
  • Predictable LLM interactions: Prompt preservation and parameter formatting issues indicate abstraction layer fragility

5. Bugs & Stability

Severity Issue Status Fix PR
🔴 Critical #2456 onboard causes system deadlock, SSH lockout CLOSED Unknown—needs verification
🔴 Critical #2443 LiteLLM supply chain compromise (credential stealer) OPEN #2485, #2491, #2494
🟡 High #2465 RoutingProvider hardcodes LiteLLM, breaks OAuth providers OPEN None identified
🟡 High #2440 openai_codex unstable prompt_cache_key (per-turn hash) OPEN None
🟡 High #2493 LangSmith broken after LiteLLM removal OPEN None
🟡 High #2476 max_tokens/max_completion_tokens mutual exclusion OPEN #2491
🟢 Medium #2208 Telegram Android message duplication CLOSED Unknown
🟢 Medium #2469 Qwen-3.5 JSON format error OPEN None
🟢 Medium #2467 WeChat group size limit exceeded OPEN None

Stability Assessment: The LiteLLM removal has created a regression cluster affecting multiple providers and integrations. The project is in a stabilization sprint with multiple competing fixes in flight.


6. Feature Requests & Roadmap Signals

Feature Source Likelihood in Next Release Rationale
/skill slash command #2489, #2488 High PR already open; addresses explicit skill activation UX gap
AWS Bedrock native provider #2485 High Critical for LiteLLM migration; PR ready
Discord.py migration #2486 Medium-High Improves maintainability; significant refactor
Nano Team Mode (multi-worker async) #2013 Medium Large feature; 2+ weeks in PR queue
Endpoint Channel (OpenAI-compatible API) #1861 Medium Enables external integrations; architectural significance
OpenClaw plugin bridge #2484 Low-Medium Ecosystem expansion; needs design discussion
Step Fun provider #2470, #2472 Merged Already complete
Sandboxed exec (bubblewrap) #1940 Medium Security hardening; addresses #2456 class of issues

Roadmap Signal: The project is prioritizing provider resilience and security hardening over net-new capabilities, likely preparing for a stability-focused release.


7. User Feedback Summary

Pain Points

Theme Evidence Severity
Supply chain anxiety LiteLLM quarantine panic; requests for dependency audit 🔴 High
Provider fragility Multiple issues on max_tokens/max_completion_tokens, model-specific errors 🔴 High
Message handling reliability Telegram duplication, WeChat limits, burst message coherence 🟡 Medium
Integration breakage LangSmith, OAuth providers failing post-LiteLLM 🟡 Medium
Misleading UX "I'm doing it" messages with no action executed #2487 🟡 Medium
Multi-user safety onboard locking shared systems 🔴 High

Use Cases Emerging

  • Enterprise multi-user deployments (Linux servers, shared environments)
  • Multi-platform messaging (Telegram, WeChat, Discord, Feishu/Lark)
  • Cost-conscious LLM usage (prompt caching, heartbeat optimization #2482)
  • Team/collaborative agent workflows (Team Mode demand)

Satisfaction Indicators

  • Active contribution (51 PRs) suggests strong community investment
  • Rapid response to Step Fun request (issue → merged PR in 24h) shows maintainer responsiveness
  • Multi-language documentation effort (#2232) indicates global user base commitment

8. Backlog Watch

Item Age Risk Action Needed
#144 Gemini API key error ~7 weeks Stale provider config issue; may be affected by LiteLLM removal Verify if still reproducible; close or migrate to new provider system
#1861 Endpoint Channel ~2 weeks Large architectural PR; blocking external integration use cases Maintainer review for merge decision
#2013 Nano Team Mode ~2 weeks Complex feature; may need rebase as codebase evolves rapidly Rebase + final review, or defer to next cycle
#1940 Sandboxed exec ~2 weeks Security-critical; addresses #2456 class of issues Priority review given safety implications
#1360 Heartbeat structured decision ~3 weeks Context budgeting improvement; builds on other PRs Dependency check + integration test

Maintainer Attention Recommended: The LiteLLM removal fallout (#2443, #2465, #2493) requires coordinated response—potentially a migration guide and compatibility matrix. The sandboxed execution PR (#1940) should be prioritized given the onboard deadlock report.

Zeroclawzeroclaw-labs/zeroclaw

ZeroClaw Project Digest — 2026-03-26

1. Today's Overview

ZeroClaw shows high development velocity with 89 total updates in the last 24 hours (39 issues, 50 PRs), indicating an active pre-release stabilization phase. The project is currently addressing critical regressions in v0.6.1-v0.6.2, particularly around WebSocket streaming, Matrix channel reliability, and memory management. No new release was cut today, suggesting maintainers are accumulating fixes for a patch release. The contributor base appears broad with multiple community PRs addressing production pain points.


2. Releases

None today. The project remains on v0.6.2 (released prior to this window).


3. Project Progress

Merged/Closed PRs (15 total, key items)

PR Description Impact
#4680 Matrix mention_only config for group room filtering Restores expected behavior for Matrix @-mention gating
#4714 Remove PostgreSQL backend, consolidate on SQLite Breaking change; ports KG features to SQLite before removal
13 additional merged/closed PRs in routine maintenance

Active Development (35 open PRs)

Critical fixes in flight:

  • #4675 / #4690: WebSocket streaming regression fix (v0.6.1 "unknown does not support streaming")
  • #4665 / #4700: Matrix multi-room listening restoration
  • #4707: Auto-generated Matrix device_id for E2EE
  • #4706 / #4709: Bounded conversation history (memory leak prevention)

Major features advancing:

  • #4166: A2A (Agent-to-Agent) protocol support — multi-agent interoperability
  • #4353: Native GitHub integration

4. Community Hot Topics

Item Activity Analysis
#3642 "Full docker image" 4 comments, 👍2 Long-standing friction: Feature-flag compilation creates barrier for non-technical users wanting WhatsApp/Matrix out-of-box. Suggests need for tiered images (slim vs full).
#4657 "Matrix channel: friction tracker" 2 comments, 👍1 Umbrella coordination issue — reveals Matrix as a problematic integration with 4+ sub-issues (E2EE, streaming, multi-room, device_id). Indicates architectural debt in channel abstraction.
#3655 "gateway web dashboard not available" 3 comments Onboarding failure: Default Docker experience broken without manual build step. Fixed by #4678 (closed).

Underlying needs: Simplified onboarding, batteries-included defaults, and Matrix reliability as a competitive requirement.


5. Bugs & Stability

Severity Issue Status Fix PR
S0 — Data loss/Security #4299 Custom OpenAI → 502 + hang after Mattermost Open None
S0 — Security #4672 Groq API key unmasked in UI Open None
S0 — Security #4478 Tool execution no-op (only shows "Used tools") Closed Fixed
S1 — Workflow blocked #4644 Web chat amnesia (2-round memory loss) Open None
S1 — Workflow blocked #4655 onboard --channels-only destroys existing config Open None
S1 — Workflow blocked #4652 install.sh --docker fails on Ubuntu/WSL Open None
S2 — Degraded #4670 All models: "unknown does not support streaming" Open #4675, #4690
S2 — Degraded #4716 MCP tools display but don't execute Open None
S2 — Degraded #4630 QQ channel WebSocket reconnection failures Open #4715
S2 — Degraded #4699 Unbounded conversation history → memory growth Open #4706, #4709

Regression pattern: v0.6.1 introduced streaming breakage; v0.6.2 Matrix refactor introduced multi-room and mention-handling regressions. Both are actively being patched.


6. Feature Requests & Roadmap Signals

Request Issue/PR Likelihood in Next Release
Full Docker image (all features) #3642 Medium — build infrastructure exists, needs CI decision
MariaDB memory backend #4668 Low — PostgreSQL just removed; SQLite prioritized
Provider-scoped fallback chains #4647 Medium — reliability config expansion
GitHub native integration #4353 High — PR active, large feature
A2A protocol support #4166 High — strategic multi-agent direction
Raspberry Pi deployment guide #4704 High — PR #4713 ready
Encrypted secrets CLI (zeroclaw secret set) #4669 Medium — security hardening

7. User Feedback Summary

Pain Points

Theme Evidence
Onboarding cliff Multiple install script failures, Docker "full image" requests, manual build requirements
Memory/scale surprises OOM on Pi, unbounded history growth — contradicts "<$10 hardware" marketing
Channel fragility Matrix dominates bug reports; QQ, Feishu, WhatsApp Web have edge cases
Configuration safety onboard --channels-only destructive behavior, secrets in plaintext

Positive Signals

  • Active self-hosting community (Mattermost, Matrix, custom OpenAI providers)
  • Demand for enterprise features (sandboxing #2972, secret management)
  • Hardware diversity: Pi 5, WSL, ARM32 all represented

8. Backlog Watch

Issue/PR Age Risk Needs
#3642 Full Docker image 11 days User churn Maintainer decision on build matrix
#2972 Plugin sandboxing 18 days Security posture Architecture review (wasm vs. process isolation)
#4166 A2A protocol 5 days Strategic direction Code review bandwidth
#4353 GitHub integration 3 days Competitive feature Final review, test coverage

Maintainer attention recommended: The confluence of Matrix reliability fixes, streaming regression patches, and A2A/GitHub feature PRs suggests a v0.6.3 patch release should be prioritized to stabilize before landing major features.

PicoClawsipeed/picoclaw

PicoClaw Project Digest — 2026-03-26


1. Today's Overview

PicoClaw demonstrates exceptional project velocity with 277 PRs and 182 issues updated in the last 24 hours, indicating a highly active development cycle. The project released v0.2.4 with focused improvements to agent error handling, logging configurability, and hot-reload documentation. The maintainer team (particularly Alix-007) is aggressively triaging with 93% issue closure rate (170/182) and 88% PR merge rate (243/277), suggesting mature CI/CD and review processes. A significant portion of activity centers on provider compatibility fixes (OpenAI, Anthropic, Gemini, Mistral, Ollama) and channel stability (Discord, Telegram, WeChat). The project appears to be in a stabilization phase post-feature rush, prioritizing reliability and multi-platform deployment experience.


2. Releases

v0.2.4 (2026-03-25)

Commit Change
14a28ae docs: note that workspace config files are hot-reloaded (#1747)
2fec249 refactor(agent): improve SubTurn error handling and logging
3293220 Add configurable logger

Assessment: Minor version bump with no breaking changes. Focuses on operational improvements: runtime configuration flexibility, better observability through configurable logging, and more robust agent sub-turn execution. Migration: seamless for existing users.


3. Project Progress

Merged/Closed PRs (High-Impact Selection)

PR Description Impact
#2016 fix(providers): improve context overflow detection and classification Critical reliability: Fixes undetected context window errors from Anthropic, ZhipuAI, GLM
#2023 feat(docker): add bootstrap init script to source-built images Deployment: Enables persistent tool installation across container restarts
#2008 feat(channels): support multi-message sending via split marker UX: LLM can now send multiple messages via <|[SPLIT]|> marker
#1998 feat(config): allow placeholder text to be string or list Flexibility: Randomized placeholder selection for dynamic responses
#2012 build(deps): upgrade pty and reorganize sqlite dependencies Maintenance: Security/stability dependency updates
#1865 Revert "feat(tools): add exec tool enhancement with background execution and PTY support" Rollback: Indicates exec tool enhancements caused regressions

Key Advancements:

  • Context management reliability — better error classification prevents silent failures
  • Container deployment maturity — bootstrap system now covers all Dockerfile variants
  • Channel expressiveness — multi-message capability improves conversational flow

4. Community Hot Topics

Most Active Issues (by engagement)

Issue Comments 👍 Core Need
#75 Local LLM/Ollama support 7 7 Accessibility: Users want cost-free, private inference
#61 File sending/receiving in chat 6 4 Capability parity: Match OpenClaw's file handling
#430 Ollama timeout (120s fixed) 7 3 Reliability: Hardcoded timeouts break local models
#148 OpenClaw SKILL compatibility 9 2 Ecosystem: Leverage existing skill marketplace
#302 Public ghcr.io container 7 2 Distribution: Easier Docker deployment

Most Active Open PRs

PR Focus Significance
#1939 ASR/TTS Refactor — voice message generation for Discord/Telegram/WeChat Major feature expansion; follows architecture from #1648
#2025 Gateway vs launcher endpoint documentation Reduces configuration confusion
#2024 Discord token persistence fix Regression fix from v0.2.4

Underlying Needs Analysis:

  • Local-first deployment (Ollama, local models) is the dominant user priority — cost, privacy, and latency concerns
  • Media handling (files, voice) represents the next feature frontier after text stability
  • Documentation clarity around deployment modes remains a friction point

5. Bugs & Stability

Severity Issue/PR Description Fix Status
🔴 High #2019 Mistral API rejects reasoning_content field (422 error) Open PR — fix pending
🔴 High #2024 Discord token updates fail after v0.2.4 (validation drops token) Open PR — regression identified
🟡 Medium #2020 Subagents don't inherit runtime tool registry Open PR — fixes #1713
🟡 Medium #2017 WeChat sync directory writability not probed Open PR — edge case handling
🟡 Medium #2022 Duplicate final reply after named-agent message tool Open PR — deduplication logic
🟢 Low #1865 Exec tool enhancement reverted (background/PTY) Merged — rolled back
🟢 Low #2021 Non-Pico channel sessions excluded from history API Open PR — data visibility fix

Stability Assessment: No critical crashes in production; issues are provider-specific incompatibilities and regression fixes from recent releases. The exec tool rollback suggests caution with privileged execution features.


6. Feature Requests & Roadmap Signals

Request Issue/PR Likelihood in Next Version Rationale
Voice/Audio (TTS/ASR) #1939 Very High Large open PR in active review; architectural foundation laid in #1648
Subagent tool inheritance #1278, #2020 High PR open, addresses autonomous workflow needs
Multi-user support #995 Medium Identified need; requires architectural changes to MEMORY.md and cronjob isolation
Telegram Forum Topics #1270 Medium Feature request with clear use case; channel expansion pattern
Self-upgrade mechanism #618 Low-Medium Acknowledged; complex cross-platform packaging
User-configurable hooks #738 Low Extensibility request; may conflict with security model

Prediction: v0.3.0 will likely ship with voice messaging and subagent tool access as headline features, with multi-user architecture beginning experimental development.


7. User Feedback Summary

Pain Points (Explicit Complaints)

Category Quote/Pattern Frequency
Configuration complexity "tried so many times to change the soul.md" — #288 Recurring
Provider setup friction Multiple "no API key configured" errors despite keys being set High
Timeout limitations 120s fixed timeout breaks local Ollama models Specific but vocal
Documentation gaps Gateway vs. launcher modes confused users Emerging
Identity persistence Docker volume/cache issues with personality files Docker users

Positive Signals

  • Active problem-solving: Users reporting issues often return to confirm fixes work
  • Deployment diversity: Windows, WSL2, Docker, Android, embedded — indicates broad appeal
  • Enterprise interest: WeChat Work, Telegram Forum Topics, multi-user requests suggest team/organization use cases

Use Case Evolution

Phase Evidence
Personal assistant CLI, Telegram, Discord individual bots
Team collaboration WeChat Work, Forum Topics, multi-user requests
Autonomous agent Subagent tool access, hooks, cronjob scheduling
Voice interaction ASR/TTS PR active testing

8. Backlog Watch

Issues Needing Maintainer Attention (>14 days, high engagement)

Issue Age Engagement Blocker
#302 Public ghcr.io container 38 days 7 comments, 2 👍 Infrastructure decision — one-click fix, policy question
#618 Self-upgrade support 32 days 5 comments, 2 👍 Roadmap prioritization — cross-platform complexity
#148 OpenClaw SKILL compatibility 40 days 9 comments, 2 👍 Ecosystem strategy — technical or business decision?
#995 Multi-user support 23 days 5 comments Architecture review — core data model changes

Stale Open PRs

PR Age Risk
#1732 Anthropic model ID normalization 8 days Low — targeted fix, may need rebase
#1657 HTTP ingress endpoint for Pico 10 days Medium — advances #850, needs API review
#1515 Cron documentation 13 days Low — docs-only, safe to merge

Recommendation: The public container registry issue (#302) is the highest-impact backlog item — minimal engineering effort, significant user experience improvement.


Project Health Score: A-

Dimension Grade Notes
Velocity A+ 450+ items/day is exceptional
Stability B+ Active regression management, some rollback needed
Community A Responsive, diverse use cases, constructive feedback
Documentation B Improving but lags feature speed
Accessibility B+ Local LLM focus addresses cost barrier

Trend: Positive — maturing from feature rush to reliable platform with expanding modality (voice) and deployment options.

NanoClawqwibitai/nanoclaw

NanoClaw Project Digest — 2026-03-26

1. Today's Overview

NanoClaw shows high development velocity with 87 total updates in the last 24 hours (37 issues, 50 PRs). The project is in active stabilization mode: 26 issues were closed against 11 new/remaining open issues, and 29 PRs merged/closed against 21 still open. No new releases were cut, suggesting the team is accumulating fixes before a version bump. Notable patterns include intensive work on CLAUDE.md initialization reliability (4 related fixes merged), Telegram platform hardening, and skill branch maintenance automation hitting repeated merge conflicts.


2. Releases

No new releases — version remains unchanged from previous cycle.


3. Project Progress

Merged/Closed PRs Today (29 total, selected highlights)

PR Description Impact
#1373 GitHub Actions workflow for Docker image publishing to GHCR Infrastructure — enables official container distribution
#1452, #1450 Task scripts agent instructions improved UX — guides agents toward schedule_task API vs. bash loops
#1346 macOS menu bar status indicator skill Platform — native macOS integration for start/stop/logs
#1349 Pre-hook gate for scheduled tasks Reliability — conditional task execution via shell hooks
#1353 Copy main CLAUDE.md template when registering main group Bugfix — fixes empty agent identity on setup
#1370 Skip bump-version/update-tokens on forks Contributor UX — eliminates workflow failures for forks
#1368 Fix claw exit code after successful output CLI reliability — graceful shutdown, proper exit codes
#1367 Default Telegram migration backfill to DMs not groups Data integrity — fixes misclassified chat types

Key advancement: The CLAUDE.md initialization saga (multiple issues since early March) appears resolved with systematic fixes ensuring agents start with proper identity contexts.


4. Community Hot Topics

Most Active by Engagement

Rank Item Type Comments 👍 Analysis
1 #783 schedule_task idempotency Closed Bug 4 0 Core API design tension — task accumulation across sessions reveals state management gaps in scheduler
2 #1075 Linux support doc mismatch Closed Docs 4 0 Platform positioning confusion — website claims Linux support, README says "coming soon"; marketing/engineering alignment needed
3 #1211 /new command for session reset Open Enhancement 3 2 Token economics pain — users hitting context limit costs; "good first issue" label suggests approachable implementation
4 #1184 K8s/Sealos deployment challenges Open Question 2 1 Production readiness gap — restricted environments expose configuration/permission assumptions

Underlying needs:

  • Session lifecycle control (#1211, #783) — users need explicit boundaries for agent state
  • Truthful platform support claims (#1075, #1445) — Linux users hitting real bugs (firewall, LID JID handling, .env corruption)
  • Enterprise deployment patterns (#1184) — air-gapped/restricted environment support

5. Bugs & Stability

Severity Issue Status Fix PR Description
Critical #1431 Closed Merged Command injection via exec() with string interpolation; race conditions in container runner
Critical #1445 Bug 1 Open None LID JID translation drops all DM messages silently — WhatsApp DMs fail for new sessions
High #1445 Bugs 2-6 Open None Firewall blocking, .env corruption, silent message drops on Linux
High #1454 Open None Sidecar exits immediately when WATCH_CONTAINER or WATCH_CONFIGS unset — async task cancellation bug
High #1451 Open None EROFS on read-only mount — watcher state file writes to wrong location
Medium #1440 Open None No intermediate feedback during long-running sessions — 10+ minute "hang" perception
Medium #1439 Open None Telegram reply-to metadata not passed to agent — breaks threaded conversations

Stability assessment: Two critical security issues addressed (#1431), but runtime reliability regressions emerging in sidecar/watcher subsystems (#1454, #1451). Linux platform showing systemic fragility (#1445).


6. Feature Requests & Roadmap Signals

Request Issue/PR Likelihood Next Release Rationale
Session reset (/new command) #1211 High "Good first issue", clear UX need, token cost impact
Persistent memory (mem0-graph) #1256 Medium Reuses existing infrastructure (Qdrant+Neo4j), mature library
Ollama MCP skill #1331 Medium Follows established skill pattern, local LLM demand
Signal channel #784 Lower "Needs Review"/"Pending Closure" status — may stall
Mattermost channel #546 Lower "Blocked" status since Feb 26, architectural alignment issues
CLI backend skill (Anthropic TOS compliance) #1343 Medium Addresses #1224 legal/compliance risk for subscription users

Signal: The project is prioritizing reliability over expansion — multiple channel PRs stalled while core fixes merge.


7. User Feedback Summary

Pain Points

Theme Evidence Severity
Silent failures #1445 LID JID drops, #1440 no progress feedback High
Setup friction #1445 Linux setup "six bugs", #1184 K8s complexity High
State accumulation costs #1211 token waste, #783 duplicate tasks Medium
Identity/persona gaps #755, #1391, #1390 missing CLAUDE.md Medium (now largely fixed)

Positive Signals

  • Appreciation for "minimalist approach" and "lightweight, secure alternative to bloated frameworks" (#1184)
  • Strong skill ecosystem contribution (Mattermost, Signal, AstrBot, macOS status bar, mem0)

8. Backlog Watch

Long-Stalled Items Needing Maintainer Attention

Item Age Status Risk
#546 Mattermost channel ~1 month "Blocked" Community contribution decay; registry pattern rewrites
#784 Signal + reactions ~3 weeks "Needs Review"/"Pending Closure" Duplicate effort risk, contributor frustration
#1049 Fix credential mounts for scheduled tasks ~2 weeks "Needs Review" Security/functional gap — scheduled tasks lack MCP credentials
#1408 claw ps subcommand 2 days Open CLI consolidation blocked on review from @​gavrielc

Skill Branch Debt

Automated merge-forward failures (#1437, #1436, #1435, etc.) indicate persistent conflicts in:

  • skill/apple-container
  • skill/native-credential-proxy
  • skill/ollama-tool
  • skill/compact

Recommendation: Dedicated skill branch maintenance sprint or archival decision needed to reduce CI noise.


Digest generated from 87 GitHub events on 2026-03-26. Data source: qwibitai/nanoclaw

NullClawnullclaw/nullclaw

NullClaw Project Digest — 2026-03-26

1. Today's Overview

NullClaw shows moderate development velocity with 3 issues and 3 PRs updated in the past 24 hours. The project is actively addressing infrastructure stability while fielding feature requests for platform expansion and multimodal capabilities. Notably, two significant PRs were closed—one improving documentation clarity and another fixing critical cron scheduler race conditions—indicating healthy maintainer responsiveness. However, no new releases were cut despite these improvements, suggesting either a pending release cycle or conservative versioning strategy. The issue backlog reveals growing user demand for Chinese market integration (WeChat) and vision capabilities, signaling where community priorities lie.


2. Releases

No new releases (v0.x.x remains current).


3. Project Progress

Merged/Closed PRs Today

PR Author Impact
#620 — docs: clarify workspace_path and system_prompt behavior vedmalex Documentation & regression testing — Resolves ambiguity from #500 by explicitly documenting that workspace_path does not suppress named agent system_prompt; adds regression test to prevent future behavioral drift
#689 — fix(cron): scheduler/worker race, memory leak, mutex lifetime, and test isolation yanggf8 Critical stability fixes — Addresses four production-risk issues: scheduler/worker race conditions, memory leaks, mutex lifetime bugs, and test isolation failures; improves daemon reliability for scheduled tasks

Technical Advancement: The cron fixes represent substantial hardening of NullClaw's job scheduling infrastructure—essential for production deployments relying on delayed/scheduled message delivery.


4. Community Hot Topics

Item Activity Analysis
#714 — Support WeChat Plugin Integration 1 comment, created 2026-03-25 Market expansion demand: User explicitly notes competing forks (OpenClaw derivatives) already support WeChat, creating competitive pressure. Massive user base in China/SE Asia cited. Underlying need: Platform diversity for regional deployment; risk of user attrition to forks
#624 — Vision Pipeline: Send images/files directly to agent 1 comment, 7-day lifespan Multimodal capability gap: User implemented as custom skill; wants native support. References sipeed/picoclaw implementation. Underlying need: First-class multimodal LLM integration without userland workarounds
#716 — Add calculator tool Open, 0 comments Tool ecosystem expansion: Comprehensive 20-operation math tool in Zig. Suggests contributor interest in expanding built-in tool suite

Engagement Pattern: Low comment volume but high strategic significance—platform coverage and multimodal support are competitive differentiators in the AI agent space.


5. Bugs & Stability

Issue Severity Status Notes
#715 — Discord scheduled DM delivery may use invalid target identifier High Open, unassigned Production risk: Scheduled/delayed DMs broken; live DMs work. Root cause: direct-session identifier reuse instead of valid DM channel target. No fix PR yet — directly related to #689's cron fixes but distinct Discord-specific channel resolution bug

Stability Assessment: #689's merge addresses underlying scheduler infrastructure, but #715 reveals a client-specific integration bug in Discord's DM channel handling that remains unpatched.


6. Feature Requests & Roadmap Signals

Feature Request Likelihood in Next Version Rationale
WeChat plugin integration #714 Medium-High Competitive parity issue; explicit mention of fork availability creates urgency
Native vision/multimodal pipeline #624 Medium Common user pain point (implemented as skill workaround); picoclaw reference provides implementation path
Expanded math/calculator tools #716 (PR) High PR already submitted; clean Zig implementation; fills gap in built-in tools

Roadmap Prediction: Calculator tool (#716) likely merges soon given PR readiness. WeChat and vision pipeline require architectural decisions on plugin system extensibility.


7. User Feedback Summary

Pain Points

  • Platform lock-in: Discord-centric design limits deployment flexibility (#714)
  • Multimodal friction: Image analysis requires custom skill development (📚 ArXiv AI Research Digest 2026-04-17 #624)
  • Scheduled messaging reliability: Discord DM scheduling appears fragile (#715)

Use Cases Emerging

  • Regional deployment: China/SE Asia market entry (WeChat)
  • Visual agent workflows: Document/image processing without preprocessing
  • Mathematical reasoning: Built-in calculation capabilities for agent work

⚠️ 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

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions