Skip to content
Pannier

Shared context for your AI agents

Change agents or devices and carry on without explaining it all again.

The agent on your phone already knows what you asked at your desk

Open another agent and say carry on. Nothing to export or paste.

Launch briefDesktop agent

Use my research and writing skills for the launch brief. Keep the tone plain.

Sections 1 and 2 are drafted in brief.md. Section 3 on pricing is next.

Saved to Pannier at 14:02

Launch briefPhone agent

Carry on with the launch brief.

Picking up at section 3, pricing. I have your research and writing skills and the plain tone.

Context from Desktop agent, 14:02

Agents save a short record, not the whole chat

As the work moves, each connected agent updates what changed. Later decisions replace older ones.

Goals
What the work is for. Ship a two-page launch brief
Instructions and preferences
How you want agents to work, once, for every agent. Plain English, short sentences
Skills
The guidance you asked an agent to use, with its version and the tools it needs. Research brief v1.2, needs web search
Decisions
Choices already made, so nobody reopens them. Postgres via Supabase
Files
The paths and links that matter right now. docs/brief.md
Progress and next steps
Where the work stands and what comes next. Section 3, pricing, is next

Your context, your rules

Every change shows which agent made it and when.

  • Choose what each agent can read

    Give your coding agent the project and your phone assistant only your preferences. Make any agent read-only, or disconnect it with one click.

  • See and correct what was saved

    Edit or delete any item. Deletes can be undone, and the full history stays with each item.

  • Nothing gets silently overwritten

    When two agents change the same thing from different starting points, both versions wait for you to pick one or combine them.

  • Take it with you

    Export everything as JSON or Markdown whenever you like.

Two versions of “Where things stand”

Current, from Phone agentSections 1 to 3 drafted. Pricing done.
Proposed, from Desktop agentSection 3 half done. Waiting on survey data.

Pick the best agent for the job. Bring your context.