Skip to content
agentusehub

Catch intent conflicts between parallel coding agents

Agents publish planned changes before editing so conflicting architecture changes across parallel worktrees get flagged before code is written.

Time saved
Not reported
Setup time
Not reported
Running cost
Not reported
Cost range
Not reported

Use case overview

Agents publish planned changes before editing so conflicting architecture changes across parallel worktrees get flagged before code is written.

Reported experience

  1. What they did

    Foremerge is a local CLI and MCP server that Claude Code, Codex or Cursor agents are wired to. Each agent registers and publishes its intent with semantic scopes such as symbol:PaymentService before editing, then claims that scope. A person installs it, runs setup, and registers the test command. The demo has one agent replacing PaymentService while another extends it.

  2. How it went

    The second agent's published intent immediately raised a HIGH replace_vs_extend advisory, with no files changed. Detection is heuristic, so it can miss synonyms or warn on compatible work. Claims only warn. No benchmark results are published.

  3. Worth knowing

    This is a pre-1.0 local-only MVP: state lives in one SQLite database that agents on one machine share, and validation commands run unsandboxed with your permissions.

Example prompts

Try with Claude Code
I run several coding agents in parallel git worktrees on [repo path]. Set up a lightweight convention where each agent writes a short planned-changes note to [shared folder or file] before editing, and write a script that compares the notes and flags conflicting architecture or file changes. Done means I can run the script and see a clear conflict report, and please show me the plan before changing any existing files.
Software & DevOps

Port a Chrome extension to Safari

The author had ChatGPT take the Session Buddy extension source from Chrome, port it to Safari and install it, taking about 20 minutes.

ChatGPT