Agent Cockpit

Agent Cockpit is an AI research agent whiteboard and MCP server that tracks evidence chains, claims, and code dependencies.

Agent Cockpit serves as a local whiteboard and Model Context Protocol server for AI research agents. It allows multiple agents like Claude Code, Cursor, and Codex to share a single board containing plans, tasks, experiments, and claims. A deterministic engine tracks content hashes down to specific JSON fields and Python functions to verify whether experimental results remain valid. When code changes silently or breaks a dependency, the system flags affected claims, blocks unauthorized commits, and raises decisions for human review. It records all actions as append-only events to enable full session replays and maintains a built-in code graph for impact analysis.

Agent Cockpit at a glance

Best for: Software engineer, Researcher
Runs on: Web, CLI, self-hosted, open source
Pricing: Free software licensed under the GNU Affero General Public License version 3.
Released: October 2026
Category: AI agent tool
Website: github.com/chutingyean/agent-cockpit

Key features

Tracks evidence chains down to individual JSON fields and recorded runs
Works with any MCP-compatible agent including Claude Code and Cursor
Enforces a single write gate to validate all agent and CLI changes
Detects unreported file changes and blocks git commits until reported
Provides a built-in code graph and rule-based code review
Manages human-in-the-loop decisions and evidence briefs
Supports append-only event sourcing for complete session replays
Runs locally on 127.0.0.1 with deterministic rules and zero LLM calls

Use it for

Coordinate multiple AI research agents on a shared whiteboard
Verify that experimental claims still hold after underlying code changes
Catch unreported file modifications before committing code
Manage human decisions and review generated evidence briefs
Inspect dependency closures and perform impact analysis on code graphs

Alternatives to Agent Cockpit

CodePilot — Consolidates multiple AI agents and local model connections in one workspace.
Agent Beacon — Retains coding agent knowledge across tools so errors do not repeat.
flightdeck — Monitor and control production and coding agents with real-time token budgets and kill switches.
Rinkata — It keeps humans and AI agents aligned on intent as work evolves.
Alook — Connects existing local coding agents to team communication channels.
Adyen — Share AI agent resources across coding tools and repositories easily.
BrowserOS — Runs entirely locally and lets AI agents use real user logins to complete tasks.
Raindrop Workshop — Provides real-time observability and self-healing evaluation loops for local AI agents.
Beeuni — Agents retain context and improve skills over time through a controlled review process.
APIStack — Connect AI agents to over 100 tools using one API key or a single MCP endpoint.
MindsHub — Pay-as-you-go access to multiple AI models with local agent integration.
Polylane — It autonomously investigates production errors and submits ready-to-merge PR fixes.

All Agent Cockpit alternatives