# ContextQ

> Durable, cross-session memory for AI coding agents. PostgreSQL + pgvector + Elasticsearch behind one API and a 16-tool hosted MCP endpoint, with a Claude Code plugin that boots each session with its prior context.

## Install

- [Agent-executable install guide](https://app.contextq.dev/install.md): the canonical install contract; every step has a stable id.
- [Install manifest](https://app.contextq.dev/.well-known/contextq-install.json): machine-readable endpoints, versions and error codes.
- [Human install page](https://contextq.dev/install): one prompt to hand to your agent.

## Docs

- [Documentation](https://contextq.dev/docs): architecture, MCP tools reference, authentication, plans and limits, self-hosting.
- [Getting started](https://contextq.dev/getting-started): manual MCP configuration for agents other than Claude Code.
- [REST API reference](https://app.contextq.dev/docs): request and response schemas.
- [Pricing](https://contextq.dev/pricing)
- [Benchmarks](https://contextq.dev/benchmarks): LongMemEval s variant, 30 questions, 70.0% (21/30), with methodology and caveats.