Company.brain

Your company's context, in one place every agent can read.

Your agent doesn't know your people, your projects, or what shipped yesterday. Stash connects every tool your company uses into one permission-aware source of truth, exposed as a filesystem any agent can read. Set up in minutes, scoped to exactly what you grant.

Connect once. Your agent reads it all.

GitHub

Pick repos your agent can navigate like a filesystem.

Google Drive

Index My Drive and read docs on demand.

Gmail

Search messages and read email on demand.

Notion

Pages and databases you share with Stash.

Slack

Channel history, kept in sync.

Granola

Meeting notes and transcripts.

Jira

Search issues across a project.

Asana

Navigate tasks from a project.

Gong

Call transcripts, kept in sync.

Agent.drive

A Drive that speaks your agent's language.

Your brain isn't just readable — it's a place agents work. Pages are real Markdown and HTML, mounted as a virtual filesystem your agent can ls, find, and rg, the same way it works in a repo. Package repeatable processes as Skills — folders with a SKILL.md — and share them with a link.

Markdown & HTML, natively

Pages are real Markdown and HTML, plus CSV, PDF, and images. The formats your agent already reads and writes, not a proprietary block format.

A virtual filesystem

The whole Drive mounts as a filesystem your agent can ls, find, and rg. Pages, files, Skills, and session transcripts sit side by side, addressed by path.

Skills are just folders

A Skill is a folder with a SKILL.md and the pages, files, and sessions it needs. No new format to learn; your agent reads it like a repo.

WYSIWYG HTML editing

When your agent builds an HTML page, a report, a dashboard, or a deck, you edit it visually in a WYSIWYG editor. Adjust the result by hand without touching the markup.

Real-time collaboration

You and your agent edit the same page at the same time, two cursors at once. Edits save automatically.

Reachable everywhere

Through the CLI, the Stash MCP server, or the HTTP API. Point Claude Code, Cursor, Codex, or OpenCode at it and they read and write directly.

Sessions land here too

Every coding-agent session streams in automatically, indexed alongside your docs. No manual upload, no copy-paste.

Connect

OAuth in and permissions are inherited automatically. No tokens to paste, no scripts to run, and access stays scoped to exactly what you grant.

Synthesize

Stash continuously maps your sources into a context graph of the company: the people, projects, and decisions, and how they connect. Kept accurate in real time, not exported once and left to go stale.

Serve

Ask one question and your agent answers across every source at once, as structured results or LLM-ready Markdown, with citations back to the source.

Give your agents the why, not just the what.

See Memory & Observability →