Trail-codeby TrailTry it on your repoGitHub11,080
Previously Graft

Your agent knows what your team already learned and decided.

Shared memory for your team’s coding agents, stored in your repo and synced with git.

Set up Trail-code in this repo:
1. Run `npm install -g @trailhq/trail-code` to install the CLI.
2. Run `trail init --agents claude` in this repo. It builds the code map and wires Trail-code into Claude Code.
3. Run `trail status` and tell me what it shows for the code map and learnings.
Then remind me to restart Claude Code so the next session picks it up.
Every agent leaves what it learns on one trail. The next one picks it up.

Mark, Sam and Nina's agents each leave a learning on one shared trail: keep retries at 4, use pnpm not npm, and no ORM in /api. Jeff's agent starts later, runs the trail, and picks up all three.

Trusted by teams at

  • Ryanair
  • Volkswagen
  • Schneider Electric
  • Juniper Networks
  • Mondelez
  • Roche
  • Bayer
  • Publicis
  • Philip Morris International
  • Topgolf

The numbers

You stop re-explaining the same thing to AI.

Same model on both sides, with and without the code map. Time and cost over 162 runs.How we measured

60%less agent time per task
32%lower token cost per task
33/50real bug fixes solved, vs 27 without the map

What it does

Every session teaches your agent.Every commit teaches your team.

01

Learns from every session

Your agent saves what it figures out: fixes, decisions, and what didn't work.

02

Shared through git

Learnings are files in your repo. Commit them and your whole team has them.

03

Works with your agent

Claude Code, Codex, Cursor, Kiro, or any other AI agent.

How it works

How the trail grows.From one session to your whole team.

  1. 01 · Set up

    Set it up in your repo

    Paste the setup prompt into your agent, inside your repo. Restart the agent once so the next session picks it up.

  2. 02 · Learn

    Your agent saves what it learns

    When your agent figures something out, it writes it down as a short markdown file in .trail/ in your repo.

  3. 03 · Commit

    You commit it like code

    Learnings go through your pull request like any other change, so your team can review them.

  4. 04 · Share

    Your teammates' agents pick it up

    On their next pull, their agents start from what you learned, and flag a change before it goes against a team decision.

  5. 05 · See

    Your team sees what changed

    A dashboard with what happened across your org in the last 24 hours, and what it means for your work.

Your dashboard

Start every morning knowing what changed.What your agents learned, and what moved across your org overnight.

Get started

Start your team's trail.Paste this into your agent, inside your repo.

Set up Trail-code in this repo:
1. Run `npm install -g @trailhq/trail-code` to install the CLI.
2. Run `trail init --agents claude` in this repo. It builds the code map and wires Trail-code into Claude Code.
3. Run `trail status` and tell me what it shows for the code map and learnings.
Then remind me to restart Claude Code so the next session picks it up.

Or run it yourself:

npm install -g @trailhq/trail-code
trail init
Claude CodeCodexCursorKiroor any other AI agent
MIT licensedNo API keyAnonymous usage stats, opt-out