Run user simulations from your coding agent.

Connect Fraser to your agent.

Choose your client, add one configuration block, then restart it.

Remote MCP server for Claude Code

  1. claude mcp add fraser \
  2. --env FRASER_API_KEY=your-token \
  3. --env FRASER_API_URL=https://fraser.lythe.ai \
  4. -- \
  5. uvx \
  6. --from \
  7. fraser-gate@latest \
  8. fraser-gate-mcp

fraser-gate — mcp — 96×24

[x] Persona cast ready

[x] Funnel report

> fraser_start_review http://localhost:3000

run_dd4b6f68 · tunnelled from localhost · 5 personas

> fraser_get_status

running · personas 3 of 5

> fraser_get_status

complete · 9 findings

> fraser_get_findings

goal_outcome friction · 3 partial · 2 blocked

→ Add a follow-upesc to stop
• Gate (shift+tab to cycle)
fraser-gate · 5 personas · 100%
/ commands · @ personas · ! shell

A review loop your agent can run.

  1. 01

    Start a review

    Point Fraser at the flow you want to test and select the personas you want to send through it.

    fraser_start_review
  2. 02

    Read the findings

    Ask for status while the review runs, then retrieve where each persona stalled, what they misread, and what they completed.

    fraser_get_findings
  3. 03

    Validate the fix

    Run the same goal and personas again after a change, so the next decision starts from comparable evidence.

    fraser_validate_fix

Scoped by design

Give your agent Fraser review tools and the environments you choose, without broad repository write, shell, or production-data access.

AVAILABLE TOOLS

fraser_start_review

fraser_get_findings

fraser_validate_fix