Your agents say they’re done.
Prove it.
For teams running coding-agent fleets who can’t trust the output.
Prismatic Engine is the auditable control plane for AI agent work. It verifies every run, signs the receipt, and gives you visibility into the black box — across all your harnesses.
AI agents are a black box. That’s the gap.
You hand work to agents. They hand back finished. Finished by whose measure? Without verification, you’re flying blind — and the bigger the fleet, the blinder.
Produce. Verify. Receipt. Merge.
One engine. Every harness. Full audit.
It runs its own development.
Every change to Prismatic Engine goes through the same machinery: verified in CI, receipted, merged under the same gates. The trust ledger, the receipts, the quarantine history — all in the public repo. We don’t ask you to trust the engine. We ask you to verify it.
Your verification shouldn’t depend on your provider.
A GitHub outage reports BLOCKED — never PASS. Verification receipts are provider-neutral: GitHub, GitLab, Bitbucket, Forgejo, or local. No mandatory cloud, no mandatory backend.
Open source, plainly.
Prismatic Engine is free and open source under AGPL-3.0-only. Use it, fork it, extend it, self-host it. If you run a modified version for others over the network, you share your source — that’s the whole deal.
Running in five steps.
$ cd prismatic-engine
$ python3 -m venv .venv
$ . .venv/bin/activate
$ python -m pip install --upgrade pip
$ python -m pip install -e ".[gateway]"
$ cp .env.example .env
$ python scripts/public_launch_smoke.py
# expected: PUBLIC_LAUNCH_SMOKE_OK
Linux or macOS · Python 3.10–3.13 · Developer preview — not a hosted service.