Skip to content
proactive-gate

proactive-gate vs Cue

Cue (murtazox04/Cue, version 0.5.4 in docs/COMPARISON.md) and proactive-gate both answer the question “may this message go to this person now”. They are built differently.

What each one is

Section titled “What each one is”
  • Cue is a self-hosted server in Python. It keeps per-person attention budgets, applies quiet hours, builds digests, holds risky messages for a reviewer, exposes an MCP server and sends messages.
  • proactive-gate is a library that runs inside your process in TypeScript, Python or Go. It decides and never sends. It needs no server.

Same inputs, measured

Section titled “Same inputs, measured”

npm run compare runs the same cases through Cue’s own policy function and through proactive-gate. The full table, the pinned versions and commits, and the run date are in docs/COMPARISON.md.

  • Is this instant quiet? proactive-gate answered correctly on 13 of 13 cases and Cue on 13 of 13. The cases include a zone at UTC+05:45, a 30-minute daylight-saving change and the repeated hour when clocks go back.
  • When may the message go? proactive-gate answered correctly on 11 of 11 cases and Cue on 10 of 11. In the one case that differs, the quiet window ends inside the hour a daylight-saving change skips. The earliest allowed time is 07:00Z on 2026-03-08; Cue answers 07:30Z, 30 minutes later than needed, by this repository’s definition of the window end.

Passing these cases says nothing about inputs that are not in them. Only Cue’s pure policy function was run, not its server.

Where Cue does more

Section titled “Where Cue does more”

Cue builds digests, holds risky messages for a reviewer and sends messages. proactive-gate does none of these. A held message goes back to your code, with nextAllowedAt when the check that held it can name the time. If you want one tool that also sends and batches, Cue fits.

Where proactive-gate fits

Section titled “Where proactive-gate fits”

You already send messages and only need the decision, you want it inside your process, you use TypeScript or Go as well as Python, or you want the same behaviour in three languages held to one spec with fixtures.

Run it yourself: npm run compare:fetch && npm run compare in a clone of the repository.