Modern AI coding agents with native RAG, Agent Client Protocol, and intelligent context management.
Meredith is an AI coding agent designed for real-world software engineering. It combines retrieval-augmented generation, smart context windows, and a plugin architecture that adapts to your workflow.
Built-in codebase indexing, symbol resolution, and semantic search keep your entire project in reach.
Agent Client Protocol support for seamless editor integration - VS Code, JetBrains, and beyond.
Intelligent context compression, budget tracking, and tier degradation keep the agent focused.
Inspired by the Meredith Effect's thermodynamic flow - cool analysis, deliberate transformation, and focused execution.
Reads your codebase, resolves dependencies, and builds a semantic index of every symbol and structure.
Decomposes tasks into ordered subtasks using flat or tree-of-thought strategies, then executes with verification.
Applies changes, runs diagnostics, and loops until the task is complete - with full traceability.
OpenAI, Claude, local Ollama, or MLX on Apple Silicon - swap providers without changing your workflow.
Read, write, edit, search, and glob files. Symbol-aware search across Python, TS, Rust, Go, and more.
Fetch documentation, search the web, and incorporate real-time information into the agent loop.
Persistent SQLite store that saves observations across sessions, filtered for security.
Loop detection, plan revision, and tier degradation keep the agent productive when things go wrong.
Every step is verified - edit sanity, diagnostics pass/fail, read efficiency - before the agent moves on.