Why Peaky Peek?
Traditional observability tools weren't built for agent-native debugging.
No external telemetry. No SaaS lock-in. Your agent data stays on your machine unless you explicitly configure cloud mode.
Captures the causal chain behind every action — reasoning steps, confidence, evidence, and chosen action — not just function calls.
Time-travel through any session. Play, pause, step, and seek to any checkpoint. Replay the exact state before a failure occurred.
| Tool | Focus | Limitation |
|---|---|---|
| LangSmith | LLM tracing | SaaS-first, no local-first option |
| OpenTelemetry | Infra observability | Not agent-decision-aware |
| Sentry | Error tracking | No reasoning-level insight |
| Peaky Peek | Agent-native debugging | ✓ Local-first & open source |
Feature Highlights
Everything you need to understand why your agent did what it did.
Navigate agent reasoning as an interactive tree. Click nodes to inspect events and trace the causal chain from policy to tool call.
Time-travel through agent execution. Play, pause, step, and seek to any point. Checkpoints ranked by restore value.
Find specific events across all sessions. Search by keyword, filter by event type, and jump directly to results.
Adaptive analysis groups similar failures. Surface highest-severity, highest-novelty events. Click a cluster to focus the timeline.
Get Started in 2 Steps
Requires Python 3.10+ and a cloned repo. No config files needed to start.
Step 1 — Install
pip install peaky-peek-serverStep 2 — Run (from repo root after cloning)
git clone https://github.com/acailic/agent_debugger
cd agent_debugger
uvicorn api.main:app --reload --port 8000
API available at http://localhost:8000 ·
Full documentation →