# Slipform > Scale your judgment. From chat threads to agent systems. Slipform is a Mac app for repeatable agent work. Turn repeatable AI work into a workflow you can run, inspect, and improve. See what needs review and decide what changes next. Keep one-offs in chat. Put work that repeats in Slipform. `Judgment` means the standards, exception paths, required proof, and human review points a person explicitly puts into a Slipform workflow. `Scale` means reusing that method across Slipform runs, comparing their records, and deciding what to revise. Slipform does not learn how you think, make approval decisions, or revise the workflow for you. A person defines the stages, decision paths, required proof, evaluations, and human review points. Approval and workflow revision remain human-owned. ## Current product - Workflow: explicit workflows preserve stages, handoffs, defined decision paths, checks, evidence requirements, and human gates. - Run: Codex work can be launched as durable runs whose activity, artifacts, evaluations, duration, token use, overrides, and review state remain inspectable. - Performance: run and stage history can be compared across scores, success, duration, token use, and overrides. - Decision: a person can Approve, Rework, or Fail with the run evidence in view. - Revision: evidence can inform a person's deliberate next workflow version; Slipform does not apply that revision autonomously. Conversational authoring, explicit workflows, Codex runs, run records and activity, artifacts, evaluations, defined review routes, performance history, token visibility, and human decisions are in the current beta. Operator is Alpha. It can prepare and coordinate bounded run actions while the Mac app is running. Operator does not approve its own work, revise a workflow, or act as an unattended approval system. Deterministic describes workflow structure, allowed transitions, required handoffs, gates, and records, not identical model output. Dynamic behavior stays inside decision paths a person defines and supported, authorized workflow-plan changes. Routes respond only through decision paths a person defines. They do not imply automatic model selection, performance-based routing, autonomous workflow mutation, token savings, or unattended approval. Codex execution is supported today and is the only supported runtime in the current beta. Broader provider and model support is Product direction and is not shipped. Claude Code support is planned and is not yet shipped. Codex is the first supported runtime, not the product category. ## Homepage walkthrough The first viewport uses an app-faithful, code-native Runs view derived from a current-beta internal Deep Analysis run. It shows the run waiting for review at Human Decision. The visualization is responsive HTML, not a product screenshot or live telemetry. One method. Every run. Define the work. Compare what happens. Decide what changes next. The homepage uses an app-faithful, code-native three-part walkthrough derived from current-beta Workflow, Performance, and Human Review states: - Workflow. Define the work. Set the stages, routes, required proof, and final human decision. - Performance. Compare the run. See where tokens accumulate, evaluations fall short, and future runs differ. - Decision. Make the call. Approve, rework, or fail with the evidence in view. The Performance state uses reviewed aggregates from one completed internal smoke run. In that run, recorded average token use was 270K per stage, Challenge used the most recorded tokens, and evaluation scores varied. Per-stage chart points and individual evaluation-bar heights remain visual placeholders until sanitized run exports replace them. This is not a benchmark, cost, savings, or customer-result claim. The same semantic component reflows on smaller screens. The Decision state reproduces the current-beta Human Review Gate structure. Its incomplete-decision finding is a local preview pending verification by the current run audit. It does not claim that Slipform automatically detected the finding or changed the workflow. ## Bring one job you keep repeating. If selected, we’ll turn it into a workflow together and inspect the first runs. The homepage collects interest only. Beta interest is reviewed in small cohorts, and joining the list does not guarantee access or timing. The hands-on exchange is conditional on selection and is not a promise of successful output or a participant deliverable. Current beta requires macOS 26 and Codex. Joining does not guarantee access or timing. No workflow description is collected by the interest form. ## Launch note [We wanted AI to do more work. Instead, we got more work to manage.](https://slipforms.ai/introducing-slipform/) explains why more agent execution can create more management work, and why the unit of work needs to move from a chat thread to an inspectable system. ## Agent Run Attestation Agent Run Attestation (ARA) is an open format for portable, signed agent-run records. It records what was authorized, what ran, where it stopped, and how the evidence can be checked independently. Official repository: https://github.com/slipforms/agent-run-attestation Native Slipform emission of signed ARA records is in progress and is not shipped in the current beta. Slipform does not claim current native ARA enforcement or signed-receipt emission. ## Current beta-request boundary - The site accepts an email address, affirmative consent, consent version `beta-access-2026-07-16`, and a server-authored consent timestamp. - The beta record does not collect a name, company, workflow description, queue position, analytics identifier, or request metadata. - A hidden company field is a bot trap. Submissions that fill it are discarded rather than retained as beta requests. - A request does not guarantee access or timing. ## Site and beta-request terms The public Terms and Privacy Policy apply to slipforms.ai and the private beta-access request list. The app will have separate pre-release terms before distribution. Slipform is operated by Kaska LLC. ## Pages - Home: https://slipforms.ai/ - Markdown homepage: https://slipforms.ai/index.md - Introducing Slipform: https://slipforms.ai/introducing-slipform/ - Launch note in Markdown: https://slipforms.ai/introducing-slipform.md - Agent Skill index: https://slipforms.ai/.well-known/agent-skills/index.json - Evaluate Slipform fit: https://slipforms.ai/.well-known/agent-skills/evaluate-slipform-fit/SKILL.md - Terms: https://slipforms.ai/terms/ - Privacy: https://slipforms.ai/privacy/ ## Integration boundary Slipforms.ai is a public product and beta-interest site. It does not currently publish a developer API catalog, OAuth or OIDC service, protected-resource metadata, remote MCP server, A2A agent, agent-commerce endpoint, or WebMCP tool. The private `POST /api/waitlist` route is the transport behind a human-consent form, not a public agent API.