Project management that reads your code.
Most project tools wait for you to fill them in. Trajan reads the code.
An open-source developer workspace where AI agents generate documentation, track progress, and manage your project — so you don't have to. Connect your GitHub repositories and Trajan's multi-agent pipeline handles the coordination work no one wants to do: writing the docs, charting the velocity, interpreting the tickets. Agent-ready by design.
Connect a repo and Trajan's agents read the code and write the docs — architecture, changelogs, plans — keeping them current as you ship.
Progress, velocity, and hotspots are derived straight from commits. No standups, no ticket-filing, no manual input.
A PM agent with live access to your codebase answers questions, interprets tickets, and resolves them when the matching commit lands.
A multi-agent pipeline on top of your repository — documentation, analytics, and project management, all derived from the code itself.
Connect a GitHub repository and Trajan's documentation agent produces a complete documentation set automatically. Claude Opus reads the actual codebase — up to 100,000 tokens of code context — and decides which documents to generate and what each should contain. Architecture overviews, technical docs, changelogs, implementation plans. When the code changes, docs regenerate. Request custom documents in plain English — onboarding guides, client handovers, security assessments. Fingerprint-based skip means no wasted AI calls.
Commit-driven analytics derived entirely from GitHub — no manual input, no ticket-filing, no standups. Five data views: summary, velocity charts, per-contributor breakdowns with activity heatmaps, active code hotspots, and AI-generated narratives. Velocity trends, bus factor warnings, day-of-week patterns — all filterable by time period and repository. Email digests deliver AI-written progress reports at each user's preferred time and frequency.
Not a chatbot with a system prompt. A conversational AI with live, agentic access to your connected GitHub repositories — it reads files, lists directories, and reasons about your actual codebase during a conversation. Pre-loaded with product context, work items, documents, and recent commit activity. Up to five tool iterations per message. Context transparency shows exactly what the agent can see.
Every data model, API, and access control layer was built so autonomous agents can operate on the same structured data as humans. Multi-agent AI pipeline — not a single prompt-and-pray system. Specialized agents for codebase analysis, documentation planning, document generation, changelog extraction, progress summarization, ticket interpretation. Row-level security on every table. Public Ticket API for external feedback with AI interpretation and auto-resolution from commits. Open source under Apache 2.0 — run it yourself or contribute.
Trajan is an open-source developer workspace where AI agents generate documentation, track progress, and manage a software project directly from its code. Connect your GitHub repositories and Trajan's multi-agent pipeline writes the docs, charts development velocity, and interprets tickets — the coordination work that would otherwise be done by hand.
Yes. Trajan is open source under the Apache 2.0 license, with source at github.com/kaminocorp/trajan.
Trajan is active and available now at trajancloud.com, with documentation at trajancloud.com/docs. Its specification lists status as Active.
Trajan is built on a Next.js 15 and React 19 frontend with an async Python FastAPI backend, PostgreSQL on Supabase with row-level security, hosted across Vercel, Fly.io, and Supabase. Its AI features run on Claude Opus, Sonnet, and Haiku, and it exposes a public Ticket API at api.trajancloud.com.
Trajan differs from conventional project tools that wait for people to file tickets and write updates: it reads the connected codebase directly. Its PM agent has live, agentic access to your GitHub repositories — reading files and reasoning about the actual code during a conversation — and external feedback arrives through a public Ticket API that an agent interprets, triages, and auto-resolves when a matching commit lands.