Flightdeck provides observability control plane for AI agents
The GitHub repository flightdeckhq/flightdeck offers a self-hosted observability and control plane for production and coding AI agents, according to its README on GitHub. The project streams every LLM call, MCP event, and tool call to a dashboard surfaced as a per-agent timeline and a live fleet-wide feed, per the repository. The README documents controls including token budgets, MCP allow/block rules, and live directives. Coding agents connect via a Claude Code plugin included in the repo; production agents use the Python package flightdeck-sensor with init() and patch() to register with the dashboard. Prerequisites listed are Docker Engine 28+ with Compose v2 and Python 3.10+, and the README provides quickstart commands and plugin install steps.
What happened
The GitHub repository flightdeckhq/flightdeck provides a self-hosted observability and control plane for production and coding AI agents, according to its README on GitHub. The project documents that every LLM call, MCP event, and tool call agents make streams to a dashboard, surfaced as both a per-agent timeline and a live fleet-wide feed.
Technical details
The README describes controls exposed by the dashboard, including setting token budgets, applying MCP allow/block rules, and sending live directives to production agents. The repo supports two integration paths: coding agents via the Claude Code plugin, and production agents via the Python package flightdeck-sensor. For the sensor path the README shows example calls to flightdeck_sensor.init(server="http://localhost:4000/ingest", token="tok_dev") and flightdeck_sensor.patch(), demonstrating use with the anthropic SDK and claude-sonnet-4-6.
Deployment notes
The quickstart lists prerequisites of Docker Engine 28+ with Compose v2 and Python 3.10+, and provides commands for cloning the repo, running make dev, and installing the plugin into a local claude session. The dev stack seeds a test token tok_dev by default. The project is self-hosted only - single deployment, single tenant per the README.
Editorial context
Tools that centralize telemetry and runtime controls for agent fleets address recurring operational needs around observability, policy enforcement, and debugging for multi-agent deployments. Teams running production agents commonly need per-call traces, budget enforcement, and live overrides; open, self-hosted control planes lower integration friction compared with bespoke logging pipelines.
What to watch
Watch for how the project handles sensitive prompt capture, retention controls, and integrations with existing logging and policy systems - those are typical gating factors for production adoption.
Scoring Rationale
A new self-hosted open-source observability and control plane for AI agent fleets, useful for teams running production or coding agents. The project addresses a real tooling gap but is a single GitHub release with no press coverage or adoption signals beyond the repo. Score reflects practical niche utility for agent-infrastructure practitioners.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
