# Agent Stash Documentation

Agent Stash is shared memory and **auto continuity** for AI coding tools. Install once with the CLI; Claude Code, Cursor, OpenCode, and Codex share decisions, progress, and session handoff.

```bash
npx @agentstash/mcp init --claude
```

The HTTP API still provides three primitives: **memory**, **logs**, and **schemas** — the CLI and MCP tools are the agent-friendly surface on top.

## Getting Started

- [Quickstart](/docs/quickstart) — Install the CLI (hooks + MCP), first `remember` / `save_progress` / `handoff`.
- [Getting Started](/docs/getting-started) — Full API walkthrough: register, store memory, write to a log, search.

## Coordination

- [Tasks](/docs/contracts) — Define shared objectives with subtasks, link execution logs, and let worker agents discover and claim work.

## Patterns

- [Progress Files](/docs/progress-files) — Use persistent memory to maintain continuity across sessions. When a session crashes or context fills, the next run picks up where the last one left off.
- [Cross-Model Coordination](/docs/cross-model-coordination) — Pass data between Claude, GPT, Gemini, and open-source models through shared memory and logs.

## Reference

- [Skill](/skill.md) — Agent SKILL.md (HTTP memory, x402 or register, no MCP required)
- [llms.txt](/llms.txt) — Short index for agents fetching this domain
- [OpenAPI Docs](/api-docs) — Interactive API reference (Swagger UI)
- [Agent Manifest](/.well-known/agent-manifest.json) — Machine-readable service capabilities
- [Pricing](/pricing-plans) — Tier limits and pricing