Put your team and its AI agents on the same page.

Claude, ChatGPT, OpenClaw, and Hermes read and update the same records through MCP.

USE HUTCH TO

  1. Run repeatable playbooksKeep instructions, files, owners, and current status together.
  2. Store and share team dataKeep research, customer records, product data, and files where everyone can find and update them.
  3. Hand work between agentsOne agent collects, another processes, and the team can see the result throughout.

Connect over MCP and say “save to Hutch.” Collections and views appear automatically.

~/hutch/demo.webm1600 × 676 · LOOP

Teams + agents on the same page.

People review, organize, and update the work in visual views. AI agents read and write the same records through MCP. No copy-paste handoffs, stale context, or work trapped in a chat.

PEOPLE / AGENTS / SHARED
WHO.01 — FOR PEOPLE

See and steer the work

Review records, change statuses, attach files, and organize collections as tables, boards, calendars, timelines, or galleries.

WHO.02 — FOR AI AGENTS

Pick up and continue the work

Claude, ChatGPT, Codex, and other MCP clients can query current context, add records, and update the work as they go.

WHO.03 — FOR EVERYONE

Leave it better for whoever is next

A decision, corrected record, or improved process becomes available to the next teammate or agent immediately.

The work stops living in separate chats

A useful answer in one person’s chat should become something the whole team can use. Save the result, the source data, or the improved process to Hutch, and every teammate and agent starts from the latest version next time.

READ / WRITE / SHARED
H
READ.001Your whole teamhumans, via web views
READ.002All their agentsover MCP, any client
READ.003Every runtoday, next week, next quarter
MEM.01

Across your team

Shared collections, orgs, and magic-link invites. What Maria’s agent saved on Wednesday, your agent can query on Thursday, with no copy-paste relay between people or vendors.

MEM.02

Across humans and agents

The same collection is a kanban board to a person and a query tool to an agent. Both read it, both write it, and neither side works from a stale copy.

MEM.03

Across runs

Context windows close. Collections don’t. Caught a missing step this run? Add it to the checklist, and every future run includes it, whether a person picks it up or an agent does.

People see a tool. Agents see structured data.

Every collection becomes a table, kanban, calendar, timeline, or gallery for the team. The same collection stays queryable and writable through MCP, so agents can do the work while people review, edit, and steer it.

05 VIEW TYPES
{ step: "draft subject lines", status: "done" }
- persona: "Ops Olivia" - role: agency owner - pain: tool sprawl
newsletter-template.md · 4.1 KB
Table01
Kanban02
Calendar03
Timeline04
Gallery05
~/todos
> "save anything we call out as a todo to hutch"A shared task list the team and its agents can update between sessions.
Screenshot of the ~/todos collection in Hutch
~/docs-debt
> "as we change the code, save any docs that need updating to hutch"A visible checklist of documentation gaps the team can work through together.
Screenshot of the ~/docs-debt collection in Hutch
~/triage
> "as we work, save any feature ideas or bugs to hutch"A triaged backlog the whole team works, between sessions and between people.
Screenshot of the ~/triage collection in Hutch
~/support-bugs
> "analyze the support tickets and save anything that points to a code issue to hutch"A bug backlog sourced from real customer pain in your support tickets.
Screenshot of the ~/support-bugs collection in Hutch

Use Hutch from

Any MCP-aware client works. Everyone on the team connects their own; one shared collection sits behind all of them.

08 SUPPORTED

Before you ask

FAQ.01

Do my teammates need to be technical?

No. People use Hutch in the browser: kanbans, calendars, tables. Only the agents speak MCP, and connecting one is a settings change in your AI app, not a code change.

FAQ.02

How do agents connect?

Hutch is an MCP server. OAuth-capable clients such as Claude and ChatGPT connect from their settings with a sign-in flow. Runtimes that do not support interactive OAuth, including OpenClaw, connect with a Hutch API token.

FAQ.03

Which AI does Hutch use?

Yours. Hutch has no bundled assistant and no AI credits to meter. It’s the shared workspace for the agents your team already uses and already pays for: Claude, ChatGPT, Codex, whatever ships next.

FAQ.04

What exactly is a collection?

A shared set of records and files: your customer personas are one, this week’s newsletter run is another. Agents create them automatically the first time they save; people see each one as a table, kanban, calendar, timeline, or gallery.

FAQ.05

Is my data locked in?

No. It’s Postgres underneath, the core is open source (AGPL), and everything is reachable over plain MCP and REST. Export it, or move to self-hosting whenever you like.

FAQ.06

What does it cost?

Self-hosting the open-source core is free, forever. Hosted plans are on the pricing page.

Create your shared AI workspace

Start with one playbook, reference library, or active project. Self-host the open-source core, or sign up and let us run it.