Reproduce the exact failure
Turn a rare production-like interleaving into a stable seed, step and trace your engineers can run again.
Deterministic verification for distributed systems
The failure happened once. Reproduce it every time.
Promtact turns race conditions, network faults and persistence failures into deterministic tests your team can replay locally, in CI and inside its own infrastructure.
OutcomeFrom incident reconstruction to a repeatable result release owners can inspect before the next change ships.
- Self-hosted
- Evidence-backed
- No required control plane
Run a real scenario
go run ./cmd/promtact simulate -config examples/leader-partition.json
Outcomes
What your team gets
Move correctness work from incident reconstruction into a repeatable engineering workflow with a result that can be shared, reviewed and rerun.
Make correctness a CI gate
Run declared faults and safety properties on every relevant change instead of relying on occasional manual fault testing.
Review evidence, not anecdotes
Keep versioned scenarios, traces and qualification results with clearly stated technical bounds.
Workflow
From protocol to reproducible evidence
Promtact focuses the campaign on the behavior that matters, then preserves any violation as an execution your team can revisit.
- 01
Connect your system
Model or adapt the protocol behavior that matters to your correctness boundary.
- 02
Define properties and faults
Describe what must never happen and which partitions, delays, losses or failures should be exercised.
- 03
Replay every violation
Return to the same seed, step and trace locally, in CI or during review.
Adoption
A controlled self-hosted verification model.
Use the Enterprise path when deployment and operations must fit a controlled self-hosted environment.
Enterprise
Operate a controlled self-hosted product
Run deterministic verification inside your own infrastructure, on a single host or Kubernetes, with controlled execution and replayable results.
- Go, Rust, C++ and Java integration
- single-host and Kubernetes deployment
- PostgreSQL-backed operations and job history
- share, replay and candidate-fix verification
Trust
Built to be checked
Promtact separates product claims from their evidence. Releases, scenarios and qualification results state which behavior was tested, on which commit, and within which bounds.
- Deterministic replay
- Bounded verification
- Signed release artifacts
- Self-hosted execution
The objective is not a blanket promise that distributed software cannot fail. It is a narrower and checkable contract: define the safety boundary, exercise declared faults, retain the exact execution and state what the resulting evidence does - and does not - establish.
Because execution remains inside the team’s environment, its scenarios and results stay available under the organization’s own access, retention and review controls.
Engineering record
Built by an upstream contributor
Review every pull request and upstream commit directly on GitHub. The record is engineering history, not a substitute for product evidence.
Review the public contribution profileStart with the hard case
Bring us the failure your current tests cannot reproduce.
Start with the protocol, failure mode and correctness property that matters most. We will determine whether Promtact fits directly or whether a scoped adapter and verification campaign is required.