Independent technical work

Nexus Code

Operational question: how can an agentic coding tool make its actions reviewable instead of hiding them behind a single permission switch?

Open source (FOSS) · solo · not a commercial product · Solo maintainer

View repositoryDiscuss a roleConnect on LinkedIn

Nexus Code is an open-source (FOSS) local coding client and harness: a TypeScript agent and sidecar, Rust workspace tools, and a terminal interface focused on reviewable agent actions.

It is aimed at individual developers who want to run the client locally, connect it to an OpenAI-compatible inference endpoint, and decide what it may do in a workspace. Documented endpoints include llama.cpp, vLLM, LM Studio, and Ollama, alongside cloud providers.

Implemented behaviour documented in the public repository includes a Ratatui TUI with REPL and one-shot modes; approval postures, workspace trust, and path controls (with documented limits); optional sandboxing and egress approval patterns described in docs; and a Linux x86-64 installer with checksum verification.

This is independent FOSS work, solo-maintained. It is not a commercial product today. A commercial offering may be built on it later; that is not claimed here. It is not multi-tenant hosting, enterprise deployment evidence, or a claim of paying customers.

Implemented

Evidence you can inspect

Approval and workspace-control artifacts from the public repository. Examples of how controls are implemented—not enterprise deployment or customer evidence.

Approval model in practice

Approval postures, workspace trust, and path controls decide what the agent may do. Optional sandboxing and egress patterns are documented with limits.

Scope boundaries