Graft (now Trail-code)

What is Graft (now Trail-code)?

Graft is an open source tool that gives AI coding agents a map of your repository and a memory of what your team has already learned. It has been renamed Trail-code by its publisher, Trail, and every graft command still works. Running init parses the repo with tree-sitter into linked Markdown nodes covering files, functions and callers, and this step needs no model, no embeddings and no API key. The second part, learnings, lets the agent save short Markdown notes in a .trail folder when it works something out or you correct it, so teammates pick them up through git and review them in pull requests like code. It is set up mainly for Claude Code and also names Codex, Cursor, Kiro, Gemini CLI, GitHub Copilot and Windsurf. Developers and engineering teams use it to stop re-explaining the same decisions to their agents. The CLI is free under the MIT license, and the vendor reports shorter agent runs and lower token cost in its own benchmark.

Use cases

• Give coding agents a map of a large repo
• Share agent learnings through git
• Review agent-written notes in pull requests
• Cut tokens spent exploring a codebase
• Check new changes against team decisions
• Set up Claude Code with persistent project memory

Who it's for

• Software developers
• Engineering teams
• Teams using AI coding agents
• Open source maintainers

Why use Graft (now Trail-code)

• Agents start from a map instead of searching the repo cold, which the vendor's benchmark ties to less time and lower token cost.
• Learnings live in the repository, so they are reviewed, versioned and shared like any other change.
• No server or sign-in is needed for the map, hooks, learnings or the check command.
• It works with several agents, so a team mixing Claude Code and Cursor can share the same notes.
• It is free and MIT licensed, which keeps adoption simple.

Graft (now Trail-code) features

• Tree-sitter code map of files, functions and callers
• Map built with no model, embeddings or API key
• Learnings saved as Markdown files in the repo
• Learnings shared through git with no server
• A check command that flags conflicts with earlier learnings
• Setup for Claude Code, Codex, Cursor and Kiro
• Existing graft commands still work after the rename
• Upgrade command that adds the trail command
• Optional deep summaries when an API key is supplied
• Anonymous usage stats that can be turned off

Graft (now Trail-code) integrations

• Claude Code
• Codex
• Cursor
• Kiro
• Gemini CLI
• GitHub Copilot
• Windsurf
• Git

How to get started with Graft (now Trail-code)

1. Install the CLI globally with npm, using the Trail-code package or the older Graft package.
2. Run init in your repo to build the code map and wire it into Claude Code.
3. Restart your agent so the next session loads the map.
4. As the agent works, it saves learnings as short Markdown files under .trail.
5. Commit those files and let teammates review them in the pull request; their agents pick them up on the next pull.

Graft (now Trail-code) limitations

• A shared team workspace and the Skills and Guardrails learning types are listed as coming soon
• It is built for coding agents working in a repo and does not remember end users across chat sessions
• Anonymous usage stats are sent unless you opt out
• Some pages and package names still say Graft while the product is now called Trail-code

Graft (now Trail-code) pricing

Graft (now Trail-code) is open source, so you can use and self-host it for free.

Prices can change and may vary by country or exclude tax. Check the official pricing page.

Graft (now Trail-code) alternatives

The main Graft (now Trail-code) alternatives listed on Debuly are Zep, Aider, Mem0, Continue, ContextStream and Augment Code. The table compares pricing and platforms; order is A–Z.

1. Ai

   Aider

   Open-source AI pair programming tool that runs in your terminal and edits code in your local git repository using the LLM you choose.

   CodingDeveloper platforms
   Open sourceVisit Aider (official site)
2. 

   Augment Code

   Coding agent platform with a codebase Context Engine, Cosmos cloud agents and the Auggie CLI, priced as flat team plans plus usage.

   CodingAutomation and agentsDeveloper platforms
   PaidVisit Augment Code (official site)
3. 

   ContextStream

   Shared project memory, code search and agent coordination for AI coding tools, connected through MCP and a browser workspace.

   CodingDeveloper platformsAutomation and agents
   FreemiumVisit ContextStream (official site)
4. 

   Continue

   Open-source AI coding assistant for VS Code and JetBrains with chat, agent, edit and autocomplete modes, now part of Cursor.

   CodingDeveloper platforms
   Open sourceVisit Continue (official site)
5. 

   Mem0

   Memory layer for AI agents and apps, available as an open-source package or a managed platform that keeps user context across sessions.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Mem0 (official site)
6. 

   Zep

   Managed memory and context platform for AI agents, built on the Graphiti knowledge graph, with Python, TypeScript and Go SDKs.

   Developer platformsAutomation and agents
   FreemiumVisit Zep (official site)

Graft (now Trail-code) FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from trailhq.com · fetched 23 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.