Inferpal
A fully local, agentic AI pair programmer for Visual Studio 2026 and VS Code that runs entirely on local models, with no API key, cloud dependency or telemetry.
Inferpal chains 28 built-in tools plus user-defined shell tools and MCP servers to read and write files, run commands, build, test and search a codebase, with every write or command behind an approval gate and a workspace sandbox; independent read-only tools run in parallel.
It offers inline ghost-text completions with several accuracy presets, background-indexed semantic codebase search that fuses cosine similarity and BM25, and a "Smart Fix Protocol" that feeds a polyglot build or typecheck's errors back to the agent after every edit. Safety features include a hard denylist for catastrophic commands and a hardened SSRF guard.
It runs against Ollama, LM Studio, or any OpenAI-compatible server such as llama.cpp or vLLM, locally or on a remote GPU host, sharing one engine across both editor extensions.