Fido

An agent-agnostic developer workflow for importing errors from Datadog, investigates them in code, and proposes fixes as draft GitLab MRs.

Install

curl -fsSL https://raw.githubusercontent.com/jorgenbs/fido/main/install.sh | sh

Or with Go:

go install github.com/jorgenbs/fido@latest

To update an existing installation:

fido upgrade

Quick Start

# Create config
mkdir -p ~/.fido
cp config.example.yml ~/.fido/config.yml
# Edit ~/.fido/config.yml with your Datadog token, services, and repository paths

# Start Fido (dashboard + background sync)
fido

Open http://localhost:8080 to access the dashboard.

Prerequisites

• Datadog Personal Access Token with error_tracking_read, error_tracking_write, apm_read, and logs_read_data scopes
• glab CLI authenticated (for MR creation)
• An AI agent CLI (default: claude)

How It Works

Fido processes errors through a pipeline of three phases. Each phase produces a markdown report in ~/.fido/reports/<issue-id>/. You can intervene, skip, or re-run any phase at any time.

Phase 1: Scan

Fido queries Datadog Error Tracking for new unresolved errors across your monitored services. For each new error it stores a report with the title, stack trace, timestamps, occurrence count, and links back to Datadog.

Phase 2: Investigate

An AI agent (Claude by default) receives the error report together with your repository and produces a root-cause analysis: affected files, suggested approach, confidence level, and relevant Datadog trace links.

Phase 3: Fix

The AI agent takes both the error and investigation reports, implements the fix on a feature branch, and opens a draft GitLab Merge Request for your team to review.

Datadog Error Tracking
        │
        ▼
   fido scan          → error.md + stacktrace enrichment
        │
        ▼
   fido investigate   → investigation.md
        │
        ▼
   fido fix           → fix.md + draft MR
        │
        ▼ (if CI fails)
   fido fix --iterate → re-fix using CI failure logs

CLI

Daemon

fido                  # start dashboard + background sync (daemonized)
fido status           # check if the daemon is running
fido stop             # stop the daemon
fido serve [--port]   # run in foreground (for development)

Error Pipeline

# Scan Datadog for new errors
fido scan [--service <name>...] [--since <duration>] [--limit <n>]

# Import a specific Datadog issue by ID
fido import <issue-id>

# Investigate an error (runs AI agent)
fido investigate <issue-id> [--service <name>]

# Fix an error (creates draft MR)
fido fix <issue-id> [--service <name>]

# Iterate on a fix when CI is failing
fido fix <issue-id> --iterate [--service <name>]

Browsing

# List tracked issues
fido list [--status scanned|investigated|fixed] [--service <name>]

# Show all reports for an issue
fido show <issue-id>

Utilities

fido version          # print version
fido upgrade          # self-update to latest release

Typical workflow

fido                             # start the daemon
fido scan                        # pull latest errors (also runs automatically)
fido list --status scanned       # see what's new
fido investigate <issue-id>      # AI root-cause analysis
fido show <issue-id>             # review the investigation
fido fix <issue-id>              # AI implements fix + draft MR
fido fix <issue-id> --iterate    # re-fix if CI is failing

Web Dashboard

The dashboard gives your team a shared view of all tracked errors and lets anyone trigger investigations or fixes without touching the CLI.

• Issue list — all tracked errors with stage indicators (scanned → investigated → fixed), service name, occurrence count, and last-seen timestamp
• Filter by stage — focus on what needs attention
• Scan Now — trigger a fresh Datadog scan from the UI
• Issue detail view — read the error report, investigation, and fix side-by-side
• One-click investigate / fix — kick off AI phases with progress streaming
• Datadog status — live issue status from Datadog (open/resolved/ignored/regressed)
• MR links — jump straight to the draft Merge Request when a fix is ready
• CI status — pipeline status badge (passed/failed/running/pending) for fixed issues
• Re-fix (CI failing) — trigger a new fix iteration directly from the UI; the agent receives CI failure logs and previous fix as context
• Ignore / unignore — dismiss noisy issues without losing them

Configuration

All configuration lives in ~/.fido/config.yml. See config.example.yml for a fully documented template.

Field	Description	Default
datadog.token	Datadog PAT	(required)
datadog.site	Datadog site domain (e.g. datadoghq.eu)	datadoghq.eu
datadog.services	Service names to monitor	[]
scan.interval	Daemon poll interval	15m
scan.since	How far back to look	24h
scan.rate_limit	Max Datadog API requests/minute	30
scan.observation_window	How long to watch resolved issues for regressions	24h
repositories.<name>.local	Local path to repo
repositories.<name>.git	Git clone URL
agent.investigate	Agent command for investigation	claude -p
agent.fix	Agent command for fixes	claude -p

Note: The site field is the Datadog site domain, not your org subdomain. If your URL is https://myorg.datadoghq.eu, set site: "datadoghq.eu".