Frequently Asked Questions
Everything you need to know about Repotoire.
What is Repotoire?
Repotoire is a local agent-processor runtime for development work. It gives development agents a small, inspectable computer with machine programs, source-pipeline evidence, precise traps, and durable state tied to git history.
Is Repotoire open source?
No. Repotoire is a commercial product in private beta. We provide access directly so teams can evaluate the local runtime, security model, licensing, and operating practices before rollout.
How is Repotoire different from a normal agent chat?
Normal agent chats rely on prose and ad hoc completion. Repotoire gives work an instruction set, register file, write buffer, worker scoreboard, verification gates, and precise traps so progress can be inspected and safely resumed.
How do I get access?
Request beta access through the contact page. Beta installs and onboarding are provided directly, including guidance for the CLI, macOS app, local binary trust, and repository setup.
What does the Machine run?
The Machine runs small programs made of explicit opcodes such as LOAD, TEST, IMPL, VERIFY, COMMIT, SPAWN, WAIT, JOIN, BNZ, TRAP, and HALT. These instructions define the stable workflow contract.
What does Repotoire track?
It tracks .machine state, events, source-pipeline evidence, worker packets, verification verdicts, git fingerprints, and traps. Git history is durable memory; the working tree is a write buffer until an explicit commit.
Does Repotoire send my code anywhere?
The core runtime is local-first and does not rely on background telemetry. Networked commands are explicit, previewed, and policy-gated; provider adapters can be configured according to the team's trust model.
Can it work with Codex, Claude, MCP, or local models?
Yes. Adapters can route work to different providers while preserving the Machine contract. The adapter can change without changing the semantics of the program, gates, evidence, or traps.
Does Repotoire work in CI/CD?
Yes, through explicit CLI commands and team-approved verification policy. Beta customers get help deciding which machine reports, gates, evidence files, and failure modes belong in CI.
What happens when evidence is stale?
Repotoire raises precise traps before unsafe writeback. A trap records the reason and restart point, so the next run can resume from a clean, inspectable state.
What is included in commercial access?
Access includes the local agent-processor runtime for development work, onboarding, architecture review, integration support, and guidance for operating machine programs, source-pipeline evidence, precise traps, and MCP adapters inside your team's workflow.
Is Repotoire a hosted scheduler or web dashboard?
No. The architecture is local-first. A website can document and support the product, but the execution model is the local Machine runtime, source pipeline, and adapter layer.