fraser-gate 0.1.0

MCP server + CLI that lets an AI coding agent run FRASER's persona swarm against a build and gate the fix pass/fail

FRASER Experience Merge-Gate — for AI coding agents

Point FRASER's persona swarm at a build, get back grounded, machine-readable experience findings, ship a fix, then re-run the identical cast to gate the fix pass | fail. FRASER is the loop's brain and eyes — it finds where real users break and verifies the fix helped. It never writes or merges your code; your agent does that.

Why this exists: a build can pass its own typecheck + unit tests and still have a user drop off. That gap — experience ≠ correctness — is what the gate catches.

The loop

your agent builds + deploys ─▶ fraser_start_review(target_url)   → {run_id, project_id, flow_id}
                                     │  (async, minutes — poll)
                              fraser_get_status(run_id) ─▶ complete
                              fraser_get_findings(run_id)
                                     │  → findings[] {finding_id, screenshot, reproduction, fix_hint} + fix_prompt
   your agent fixes + redeploys ◀────┘   (FRASER never touches code)
                                     ▼
                              fraser_validate_fix(project_id, flow_id, baseline_run_id, new_url)
                              fraser_get_validation(run_id, baseline_run_id)
                                     │  → {gate: pass|fail, per-finding {fixed|still_present|regressed}, new_findings}
                              gate=fail ⇒ don't merge · gate=pass ⇒ ship

Setup (one paste)

Get an API token from the FRASER dashboard (Settings → Connect MCP → Generate token), then add to your agent's .mcp.json:

{
  "mcpServers": {
    "fraser": {
      "command": "uvx",
      "args": ["--from", "fraser-gate", "fraser-gate-mcp"],
      "env": { "FRASER_API_KEY": "<your token>", "FRASER_API_URL": "https://fraser.lythe.ai" }
    }
  }
}

That's it. localhost just works — if you point a review at http://localhost:3000, the MCP opens an ephemeral public tunnel automatically so the hosted engine can reach your machine; you set nothing up.

Tools (6)

Tool	Use
fraser_start_review(target_url, mode?, objective?, context?, num_testers?, auth_notes?)	Start a review. Async — returns a run_id.
fraser_get_status(run_id)	Poll until terminal (complete/failed/timed_out). Runs take minutes.
fraser_get_findings(run_id)	Scrubbed findings + fix_prompt. You fix from these.
fraser_validate_fix(project_id, flow_id, baseline_run_id, target_url)	Re-run the identical cast on the fix. Async.
fraser_get_validation(run_id, baseline_run_id)	The pass/fail gate verdict.
fraser_cancel(run_id)	Stop a run.

It takes minutes. FRASER generates personas from your app, then drives a real browser as each one. Start the review, do other work, and poll fraser_get_status every few seconds.

Knobs (defaults are good): mode = visual (experience critique, default) or battle (defect sweep); num_testers 0 = the grounded cast; context = one line on what the product is (anchors the personas for a new/thin app); auth_notes = app login creds if it's behind your own login (secret).

CLI (same loop, for CI without MCP)

export FRASER_API_KEY=<token> FRASER_API_URL=https://fraser.lythe.ai
fraser review https://myapp-pr42.vercel.app --wait          # prints findings JSON
fraser validate <project_id> <flow_id> <baseline_run_id> <new_url> --wait   # exits non-zero on a FAIL gate

Local dev of this package: pip install -e . then run python mcp_server.py (stdio) or python cli.py.