Immutable revisions
Natural-language changes create a new reviewed version. Prior runs remain intact.
LOCAL-FIRST AI WORKFORCE
Repeatable AI work for a one-person company.
Describe a goal. Review the team and workflow. Run it again with version history, approvals, files, and evidence - on your own Mac.
macOS 14+ candidate / Python 3.12 / Apache-2.0 / package validation in progress
Your work stays local. OpsWitness is a local Mac application, not a hosted SaaS. This public website contains product information, source, release status, and feedback links only.
CURRENT AVAILABILITY
The public preview is open. The install package is not posted yet.
The Community Alpha release candidate is still completing its reproducibility and durability checks. You can inspect the source and product today. A verified wheel, checksums, SBOM, and attestation will appear together on GitHub Releases when the gate passes.
Check release statusONE SIMPLE LOOP
From a goal to work you can run again.
OpsWitness sits above your AI runtimes. It makes the team, process, decisions, and results visible without asking you to operate the underlying control planes.
-
1
Describe the outcome
Start from a goal, a proven template, or an earlier planning conversation.
-
2
Review the AI team
See roles, reporting lines, stages, model choices, and human checkpoints.
-
3
Confirm and run
Execution starts only after an immutable plan version is reviewed and confirmed.
-
4
Reuse what worked
Open readable results, inspect evidence, rerun, fork, or revise the architecture.
THE ACTUAL PRODUCT
A calm workspace, not another control panel.
These are real OpsWitness screens using synthetic demonstration data.
BUILT FOR REPEATABILITY
The process gets better without erasing its past.
Run history
Each execution keeps its process, inputs, decisions, results, and status.
Evidence-bound files
Artifacts are content-addressed and linked to the run that produced them.
Human control
Use automatic mode for routine work or require approval at sensitive checkpoints.
Repeatable Work
Prepare a proven team and workflow in one click, then review before dispatch.
Reviewed memory
Agents propose lessons; only approved, versioned memory can affect new planning.
HONEST ALPHA BOUNDARIES
Useful now, explicit about what is not ready.
Included in the Alpha candidate
- Local single-operator Workspace and Work UI
- Plan review, immutable revisions, rerun, and fork
- Team execution with input and approval checkpoints
- Pause, continue, stop, History, Results, ledger, and CAS
- Built-in repeatable-work templates and reviewed Workspace Memory
Not promised in this release
- Stable durability, SLA, or unattended critical operations
- Hosted SaaS, multi-user accounts, or a mobile app
- OpenClaw, Work-calling-Work, or silent autonomous learning
- Tax filing, customs filing, insurance advice, or other licensed decisions
- DeepSeek or Grok execution adapters
- A downloadable package before the release evidence gate passes
COMMUNITY ALPHA AVAILABILITY
One verified package, when the evidence is ready.
OpsWitness will be distributed through a GitHub prerelease, not PyPI. Until that prerelease exists, this page will not offer an unverified download.
What the release will include
- Python wheel and source archive built from the tagged commit
- SHA-256 checksums and a machine-readable build manifest
- SPDX software bill of materials and GitHub attestation
- A blank-install smoke test before the download link goes live
The first package targets macOS 14+ with Python 3.12. Start with synthetic or non-critical work; durability remains experimental in Alpha.
Release candidate validation in progress
The download button will appear only after the exact public build passes the release checks. No email signup or hosted account is required.
HELP SHAPE THE ALPHA
Tell us which repeatable process should be easiest first.
Before the package is posted, you can review the screens and source or request a workflow. After release, the most useful reports describe the goal, point of friction, and whether plan, history, and final result were easy to find.