For work that spans AI sessions

Pick up where
the work actually
left off.

mcue keeps what happened, what you decided, and what comes next in one record. Your next session, in any app you connect, starts from it.

Invite-only private beta · Browser or CLI

A handoff, made visibleExample / 01
01 / ClaudeSession ends

Five interviews.
One decision.

Focus the pilot on small teams. Pricing still needs testing.

Closeout savedWritten by the agent
02 / ChatGPTNext day

“Resume the pilot research.”

Start with the three pricing questions for your next two interviews.

Resume atresearch/interviews.md
Inspect the saved record
Objective
Choose the first customer segment for the pilot.
Outcome
Reviewed five interviews. Small teams share the clearest need.
Decision
Focus the pilot on teams of two to five.
Blocker
Pricing expectations are still untested.
Next action
Draft three pricing questions for the next two interviews.
Resume at
research/interviews.md · Pilot brief

Illustrative data · agent-authored · both clients connected to mcue

Built with the work it remembers. mcue’s site, hosted service, and CLI are built using its own closeouts and resume packets.

Follow the changes

The conversation ends.
The work goes on.

A closeout preserves what happened. The saved state gives the next session a bounded resume packet.

  1. 01

    Finish a session

    Save the outcome, decisions, and next action.

  2. 02

    Inspect the record

    See what changed and who wrote it.

  3. 03

    Resume anywhere

    Continue in another connected client.

Customer research PilotIllustrative record · example data
History · every closeout kept
  1. Reviewed five interviews. Chose teams of two to five for the pilot.

    Written by an agent · hosted MCP
  2. Wrote the pilot brief and the interview guide.

    Written by Alex · CLI
  3. Shortlisted three candidate segments.

    Written by an agent · hosted MCP
Where it stands now

Next actionDraft three pricing questions for the next two interviews.

Resume at
research/interviews.md
Blocker
Pricing expectations are still untested.
  • ClarityHigh
  • Resume costLow
  • FreshnessFresh
  • TrustMedium
The next session starts from where the work stands now. The history stays there to inspect.How the record works
What reaches the next session?
  1. CloseoutThe authored account of a work session.
  2. CheckpointAn immutable entry in the lane’s history.
  3. Lane stateThe current objective, blockers, and next action.
  4. Resume packetBounded context derived from that state.

The next session reads the packet. The full record remains available to inspect.

Connect once.
Carry on.

Choose your AI app. Connect it to mcue, sign in, and let it read the work before continuing.

Hosted MCP endpointhttps://api.mcue.ai/mcp
Your AI app
  1. Open Customize → Connectors

    Select “+ Add”, then “Add custom connector”.

  2. Name it mcue and paste the endpoint

    Continue, review the detected sign-in settings, and select “Add”. Sign in to mcue when Claude asks.

  3. Turn it on in a chat

    Use the + button, then Connectors, and switch mcue on. Ask Claude to read your mcue projects.

For Claude on the web and desktop. The Free plan allows one custom connector. On Team and Enterprise plans an owner adds it under Organization settings first, then each member connects.

Hosted connections need beta access. Sign-in happens in your browser; no token to copy by hand.

Prefer local? Use the CLI

Agents change.
Your work stays.

Your record outlives any one model or conversation. Keep it on your machine, or host it for your connected apps.

On your machineLocal
~/.mcue/projects/customer-research/
  • project_state.md
  • decisions.log
  • lanes/pilot/lane_state.md
Plain files · CLI + local MCP

Complete, even offline.

Readable files you can inspect, diff, and repair. Once installed, no account is needed for local work.

What’s included
  • Full state model and reconciliation
  • Local MCP for compatible agents
  • Optional BYO-S3 remote path
  • CLI licensed Apache-2.0
Available to your appsHosted
AI appsBrowserCLI
One saved record
Available while your computer is off

Ready in your next client.

Create a project in the browser or publish a local one. Each authorized client can reach the hosted state.

What’s included
  • Persistent hosted MCP and accepted event ordering
  • Lane-level sharing, read-only by default
  • Per-client access you can restrict or revoke
  • Invite-only private beta · proprietary service
Each save is explicit
A closeout gives a work session a clear write boundary.
Authorship is visible
See whether a person or an agent wrote each closeout.
Access is yours to change
Restrict a connected client to read, or revoke its access.

Your source code stays where it is. Read the mcue thesis

A clear way
back in.

The same closeout and resume loop, from your terminal. Your beta invite includes a personal CLI download key.

TerminalmacOS · Linux

Install with the key from your invite

curl -fsSL https://mcue.ai/install.sh | MCUE_DOWNLOAD_KEY=<your-key> sh

Create a project

mcue init payments-api

Save the outcome and next action

mcue closeout payments-api

Return with a bounded resume packet

mcue resume payments-api
Plain Markdown + YAML. A record you can read.
Working with a coding agent?Have it read the state before the code.

Run mcue agents-snippet and add the access rules to your repository’s AGENTS.md or CLAUDE.md. They tell the agent to recover context first and propose a closeout when the work is done.

Agent setup guide ↗Full access rules ↗
A few useful details

Before you
begin.

Do I need both the browser and the CLI?

No. Start in the browser and connect your AI app, or install the CLI and work locally. Once installed, the CLI works offline without an account. You can connect the two later.

Can I try it without an invite?

Not yet. During the private beta, one invite gives you an account for the browser and connected apps, plus a personal key for the CLI download.

How is this different from chat memory?

mcue keeps an explicit work record: objective, outcome, decisions, blockers, and next action. A resume packet is derived from that saved state. You and your connected clients can inspect the record across sessions. It works alongside your AI app’s memory.

Does an agent need approval for every save?

An AI app you connect can write closeouts to your projects. The mcue access rules ask agents to propose a closeout and wait for your confirmation. That is an agent instruction, not an enforced approval prompt. You can restrict a connected client to read access or revoke it in your account.

Does mcue store my code or replace my project tools?

No. Your code, documents, and tasks stay in their existing tools. mcue stores the state of the work and pointers to what the next session needs. Projects can also begin without a repository.

The cue to resume

Leave the next session
somewhere to start.

Request beta access

One invite. Browser access and a CLI download key.