Renitor: your coding task outlasts any one AI agent
A published VS Code extension: when one AI coding agent stops from a quota cap, a crash, or a bad answer, another continues with the task intact, and a deterministic safety layer reviews risky commands before they run.
Independent product: designed, built, shipped
A shipped developer product: 11-agent task handoffs, automatic model failover, and a deterministic command-safety layer, local-first with zero telemetry.
Developers who use AI coding agents live with a specific frustration: a subscription caps or a session dies mid-task, and the next tool starts from zero. The repo context, the decisions, and the failing test all get re-explained by hand. Nothing on the market carried a task across agents cleanly.
- I designed, built, and published Renitor (v1.75 on the VS Code marketplace, live at renitor.com).
- Repo-aware handoffs carry an unfinished task across 11 coding agents, including Claude Code, Codex, and Cursor, as a structured checkpoint plus a paste-ready continuation prompt, with a round-trip handback that re-injects natively into Claude Code.
- A multi-model chat runs Claude, GPT, Gemini, and 22 free models in one thread, failing over automatically when a provider hits its cap.
- A deterministic safety layer reviews risky commands before they run, and a failover dry-run plus a privacy inspector show what will happen before it does.
- Local-first: no backend, zero telemetry, keys in the OS keychain.
A working, published product with real users, and the engineering to back its claims: the safety layer is deterministic rather than model-judged, the privacy story is inspectable, and the handoff format is versioned and tested. Free during the open launch.
Every capped session stays a restart from zero. The developer re-explains the repo, the decisions, and the failing test to the next agent by hand, or pays for capacity they only need in bursts.
Read the full story
Renitor started from a frustration I had myself. An AI coding agent dies mid-task from a quota cap, a crash, or a bad answer, and the next tool starts from nothing. The repo context, the decisions made, the failing test: all re-explained by hand, every time.
The core of the product is the handoff, a structured repo-aware checkpoint covering git state, decisions, what’s broken, and what’s next, plus a paste-ready continuation prompt that works across 11 coding agents. The round trip matters as much as the departure. Work done elsewhere comes back into Claude Code natively, so the original session knows what changed.
Around that core grew a multi-model workbench: Claude, GPT, Gemini, and 22 free models in one conversation, with automatic failover when a provider hits its usage cap. The same thread simply continues on the next model. Boss & Worker and Debate modes put models to work together on one task.
Two design commitments shaped everything. First, safety is deterministic. A rules-based layer reviews risky commands before they run, and a failover dry-run plus a privacy inspector show exactly what will happen before it does, so nothing depends on a model probably catching it. Second, local-first. There is no backend, no analytics, and zero telemetry, and API keys live in the OS keychain. The extension’s own infrastructure never sees your code.
It ships as a published VS Code extension, currently v1.75, free during the open launch, with the source, the versioned handoff format, and the test suite to back the claims.