- Who it helps
- Developers building APIs and MCP tools for personal assistants.
- Availability
- Version 0.2.1 was released October 22. This alpha tool uses Python 3.12 and uv; browser tests also require Node 20 and Chromium.
Check the outcome
A test defines starting records, permitted approvals and expected service changes. Saved interaction evidence supports replay and comparison, including duplicate charges or cross-account changes.
Understand the surrogate
The launch describes a Muse/Instinct-like test agent with assumptions labeled documented, unofficial, assumed or unknown. Passing this simulation does not prove compatibility with either commercial assistant.
What you can do
- For an initial evaluation, define one synthetic record, one allowed action and one forbidden change. Require the service record and final answer to agree. Open reference ↗
Where the evidence stops
The browser lane does not itself provide operating-system isolation. No CATT test was run here. One interested HN reply is not an independent reproduction.
Sources read. Original sources read October 23, 2026 UTC. No external project, model inference or reported experiment was executed.
Sources & further reading
GitHub · implementation
Consumer Agent Testing Tool README ↗
Read alpha prerequisites, service-state assertions, evidence labels and browser-lane limitations. The surrogate is not affiliated with Muse or Instinct and is not a compatibility certification.
Coverage-Cat · page checked 2026-10-23
GitHub · implementation
CATT changelog ↗
Read the dated October 20 initial release and October 22 0.2/0.2.1 entries. No release code executed.
Coverage-Cat · page checked 2026-10-23
More sources reviewed (1)
Hacker News · discussion
Show HN: CATT 0.2.1 ↗
Native launch post and one independent reply read. The author explains reproducibility limits and surrogate assumptions; the reply expresses interest rather than a completed test.
botacode · page checked 2026-10-23 · Published 2026-10-22
Coverage note: No independent end-to-end reproduction or complete video demonstration was verified in this bounded sample.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
CATT records an assistant’s tool calls and checks the resulting service state. It helps connector developers catch cases where a confident final answer hides an incomplete or unauthorized action.