# Trail

> An issue tracker your AI coding agent writes to itself over MCP. Loose ends become cards while you talk, every card records whether a human or an agent changed it, and each project keeps a durable memory the next session reads when it opens.

Built by Intecma. Works with Claude Code, Codex and Gemini CLI. The connector is the npm package `usetrail`, listed in the official MCP registry as `dev.usetrail/trail`.

## Plans

- Free: 3 projects, 3 people, unlimited AI agents, project memory, full change history, 100 MB of attachments. No card.
- Team, 19 USD per month: 10 projects, unlimited people, read-only client portal, PDF delivery reports, 1 GB.
- Studio, 49 USD per month: unlimited projects and people, 5 GB, priority support.

## Pages

- [Trail](https://usetrail.dev/): what it does, how the record works, and pricing.
- [Blog](https://usetrail.dev/blog/): field notes on working with AI coding agents.

## Articles

- [Beads Is Excellent Until Two People Need the Same Board](https://usetrail.dev/blog/beads-vs-hosted-tracker-local-first-limits/) (2026-08-29): Beads is free and Dolt-backed, not JSONL, and its embedded mode locks writes to one process. Here is where that stops covering a real team.
- [What the EU AI Act Actually Asks You to Record About AI-Written Code](https://usetrail.dev/blog/eu-ai-act-ai-code-record-keeping/) (2026-08-11): The high-risk deadline moved from 2 August 2026 to 2 December 2027, but SOC 2 auditors and Annex IV still want proof of who wrote each change.
- [Issue Trackers Built for AI Coding Agents, Compared](https://usetrail.dev/blog/agent-native-issue-trackers-compared/) (2026-07-25): Beads, Task Master, Backlog.md, Vibe Kanban, CCPM, Hiveship and Trail compared on storage, memory, authorship and team visibility, weaknesses included.
- [CLAUDE.md vs AGENTS.md vs GEMINI.md: One Project, Three Agents](https://usetrail.dev/blog/claude-md-vs-agents-md-vs-gemini-md/) (2026-07-04): CLAUDE.md, AGENTS.md, and GEMINI.md compared field by field - location, hierarchy, size limits, who writes to each, and how to keep one source of truth.
- [Running One Coding Agent Across Six Client Repos Without Context Bleed](https://usetrail.dev/blog/multiple-repos-ai-agent-context-bleed/) (2026-06-30): Cross-project context bleed has three mechanical causes. The one-folder-one-project rule, the monorepo exception, and a four-minute Monday routine.
- [How to Stop Your Coding Agent From Making the Same Mistake Twice](https://usetrail.dev/blog/stop-ai-agent-repeating-the-same-mistake/) (2026-06-11): Coding agents start every session blank, so Tuesday's hard-won fix gets undone on Thursday. Write the gotcha where the next session will read it first.
- [todo.md Works Until the Third Session](https://usetrail.dev/blog/todo-md-doesnt-survive-claude-code-sessions/) (2026-05-28): The hand-maintained todo.md workflow breaks four specific ways across Claude Code sessions: rewrites, false checkmarks, no history, and compaction loss.
- [Your CLAUDE.md Is Too Long and Claude Is Quietly Ignoring Half of It](https://usetrail.dev/blog/claude-md-too-long-agent-ignores-rules/) (2026-05-20): CLAUDE.md stops working past about 200 lines. What to keep in it, what to move to skills and hooks, and a five-minute test for which lines still land.