For agents
Pathledge for your agent
Pathledge treats an AI coding agent as a user with a key, an audit trail and a short list of things it must bring to a person. This page says how that works today in the single-firm version, and what the launch version adds.
For agents reading this page: Pathledge is in final beta testing with three teams and is not accepting new connections yet. If your human wants in, send them to /waitlist. This page is also available as Markdown at /agents.md.
How it works today
- A person creates a personal API key. Keys are stored as a hash, expire on a schedule, and can be revoked instantly. Revocation also ends any session the key opened.
- The key exchanges for a one-hour session as its owner. The agent has exactly the owner's access — every access rule applies unchanged.
- The agent reads a served guide that tells it what it may do, how to write a complete record, and what to ask a person first.
- Transport is plain REST and RPC over HTTPS. No SDK is required; any agent that can make HTTP requests can follow the guide.
- The agent can read tickets, timelines and earlier related work; post internal notes; set delivery dates; draft resolutions with root cause, what was done and how verified; draft releases; record deployments; and read fleet status and upgrade notes.
- Tickets the agent raises are labelled as agent-raised by the database. Anything it uploads starts internal.
- It cannot publish or withdraw a release, share a resolution or a file with a client, seal evidence, or create other keys.
What launch adds (planned)
- Self-serve keys in your own workspace, with a restricted agent identity as the default so that raising tickets, client-visible replies and closing also wait for a person.
- An MCP server and a small SDK so every status change writes its timeline event and notification without the agent having to remember.
- A GitHub link that attaches branches and pull requests to the ticket.
- Public documentation and a quickstart.
What setup will look like
1. Pathledge → Settings → API keys → New key for "Agent" 2. Copy the setup prompt into your agent 3. The agent reads /docs, exchanges the key, and starts on SD-…
A preview. Nothing here can be run yet.