Tramya

Local-first memory for AI coding agents

Tramya remembers the decisions, files and plan of each project on your machine — and hands that context from one AI agent to the next. Switch between Claude Code, Codex, Cursor, Gemini CLI, Cline and Copilot without re-explaining anything.

Website · Docs · Pricing · Integrations

The problem

AI coding agents forget everything between sessions. Every agent keeps its own history, in its own format, on its own island — Claude Code in ~/.claude, Cursor per workspace, Codex somewhere else. None of them read each other.

So every time a session hits the token limit, or you close a terminal, or you switch from one tool to another, you re-type the same paragraph: "we're migrating auth, the decision doc is in /docs/adr-014, don't touch the legacy path, tests run with…". You keep the code, but you lose the why.

Related MCP server: devrecall

What Tramya does

Tramya is a companion, not a replacement. It doesn't try to be your IDE or your agent — it's the memory layer that sits next to whatever agents you already use.

• 🧠 Indexes what's already there. Detects the local histories of your installed agents and indexes each session by repo, branch and Git worktree — read-only, in a few seconds, no per-project setup.

• 🔎 Search across all agents at once. One query surfaces the relevant Claude Code discussion and the Codex task and the project note, each with its provenance.

• 🤝 Hands off a context capsule. When you switch agents, Tramya generates a capsule — durable decisions, touched files, CLAUDE.md rules, current Git state, next verifiable steps — and the next agent resumes exactly where the last one stopped.

• 🔒 Local by default. Everything stays on disk. Optional encrypted Cloud sync shares project context between teammates — never raw prompts.

Supported agents

Claude Code · Codex · Cursor · Cline · Gemini CLI · GitHub Copilot · Windsurf · Aider · Hermes · OpenCode

Context is exposed to any agent that speaks MCP, plus a public HTTP API.

How it works

flowchart LR
    T["Tramya<br/>local-first memory<br/>repo, branch, worktree"]
    CC[Claude Code] --> T
    CU[Cursor] --> T
    CX[Codex] --> T
    GM[Gemini CLI] --> T
    T -->|"context capsule:<br/>decisions, files, next steps"| N["Next agent resumes<br/>where the last one stopped"]

Each agent keeps writing its own local history; Tramya indexes all of it into one searchable memory (provenance preserved) and hands the next agent a context capsule over MCP.

Install

Download the app for your platform — free tier is permanent:

👉 tramya.com — macOS · Windows · Linux

Then open a repo: Tramya detects your agents' local histories and indexes them automatically. See the documentation for the full setup, and the Claude Code integration guide to query memory over MCP.

Use it from your agent (MCP)

Once connected during onboarding, your agent can query Tramya over MCP — for example, ask for a project's context or search a past decision:

// The agent calls the Tramya MCP server, e.g.
{ "tool": "get_project_context", "args": { "project": "my-repo", "query": "auth migration decision" } }

The context comes back with provenance (which agent, which session), so answers stay grounded. Full tool list and the public HTTP API are in the docs.

Learn more

Pricing

A permanent Free tier for local use, plus paid plans (Pro, Cloud, Team) for encrypted sync and team features. See tramya.com/en/pricing.

Feedback & issues

Found a bug or have a feature request? Open an issue in this repo, or reach us from tramya.com. We read everything.