Use case · Builders

How to share project context between Claude Code, Codex, and Cursor

Each coding agent remembers its own conversation. Switching tools means manually rebuilding the brief, and important constraints get compressed or lost along the way.

The setup

A simple way to run it in Hark

  1. 01Connect every client to one hub

    Point Claude Code, Codex, Cursor, and other MCP clients at the same Hark workspace.

  2. 02Read the brief at session start

    Use get_agent_brief and start_session so every agent begins with the current project and Current state.

  3. 03Write decisions as they happen

    Record durable constraints and rationale in the journal instead of leaving them in one chat transcript.

  4. 04Save the handoff

    Call end_session with what changed and what comes next so the following client can resume the work.

  5. 05Rely on the guardrails

    Sessions are locked to one project so a client cannot read or write another one by accident, a session that goes quiet has its handoff drafted automatically, and repo export commits the same brief into your repository as .hark/HANDOFF.md and .hark/DECISIONS.md.

Starter build

Hark’s public project page

Starter builds are coming. Until then, use Hark’s own documented build as the example structure for this setup.

View the public project page placeholder →
Use this setup

Common questions

Can I use this setup for an existing AI-assisted project?
Yes. Import a conversation or brain dump, review the drafted project, and connect the repository when you are ready. You do not have to start from scratch.
Do all collaborators and agents see the same information?
They work from the same Hark project and Current state, subject to the workspace and project permissions you choose.
Can I try this without adding a card?
Yes. Hark starts with a 14-day card-free trial, and your work remains readable if the trial ends before you choose a plan.