fastctx
Fast, context-efficient repository tools for AI agents (MCP)
What is it?
What it is
A local Rust tool runtime providing structured repository operations (file read/search, file discovery, batch replacement, Bash command execution) for AI agents to reduce context overhead.
Why it exists
To minimize AI agents' focus on tool mechanics (e.g., command syntax, path handling) by abstracting repository interactions into structured input/output formats.
Who should use it
AI agent developers, LLM-based coding assistants, and engineers using MCP-compatible clients (like ChatGPT or Codex) who need efficient, structured repository interaction tools.
Who should avoid it
Users who require a purely sandboxed/web-only environment without local system access, as FastCtx requires local execution and can perform file writes and shell commands.
How it works
A quick walkthrough in plain English
How fastctx works
Step 1 of 3
You interact with it
Open fastctx, send a request, or connect it to your stack.
Features
Advantages
- Reduces model reasoning overhead by providing structured, predictable tool inputs/outputs
- Improves context efficiency by handling complex command construction and escaping internally
- Provides stable, persistent tool execution via a dedicated runtime process
- Offers robust job management with rolling output buffers and persistent storage
- Ensures safe file modifications through atomic replacement and dry-run capabilities
- Simplifies complex repository navigation for AI agents
Disadvantages
- Requires manual enablement for Bash/shell execution tools for security
- Background job output is limited to an 8 MiB rolling buffer
- Higher output tiers consume model context faster
- Requires Node.js 18+ for npm installation
Installation
native
Install with npm: ```console npm install --global fastctx fastctx ``` If the install returns 404: ```console npm install --global fastctx --registry=https://registry.npmjs.org/ ``` One-off run: ```console npx fastctx ``` Other distribution channels: ```console cargo install fastctx --locked ``` GitHub Releases provides a zip archive for Windows x64 and executable-preserving tar.gz archives for Linux x64, macOS x64, and macOS arm64.
FAQ
How do I install FastCtx and launch the control terminal?
Install via npm with a recent Node.js (≥18): ```bash npm install --global fastctx fastctx ``` The first launch opens a full‑screen TUI where you can configure settings, enable Bash tools, and start ChatGPT/Codex sessions. If you prefer a one‑off run, use `npx fastctx`.
What core tools does FastCtx provide and how can I enable the Bash command tools?
FastCtx ships with four file‑centric tools: - `read` – read text, images, PDFs, or raw bytes - `grep` – regex search in files or trees - `glob` – file pattern matching - `replace` – mechanical batch replacements The five Bash‑related tools (`run`, `run_background`, `job_output`, `job_kill`, `job_list`) are disabled by default. Enable them in the TUI under **Bash terminal** or by starting the server with `--enable-shell`.
How can I run a long‑running command in the background and later fetch its output?
Use `run_background` to start the job: ```json {"command": "cargo test", "timeout_ms": 180000} ``` It returns a `job_id`. Retrieve incremental output with `job_output` (use `after_seq` to page) and terminate the job with `job_kill` if needed. Completed jobs are listed by `job_list` and their output is retained until the storage limit is reached.
How do I control the CPU parallelism used by `grep` and `glob`?
In the TUI’s **Config → Search** panel, set `max_cpu_cores` to an integer (1‑16) or `auto` for automatic detection. The setting is written to `~/.fastctx/config.toml` under `[search]`. It takes effect after restarting the server and does not require Apply.
What non‑interactive commands can I use to manage FastCtx?
FastCtx offers a CLI for scripted use: - `fastctx apply --tier standard --yes` – install and configure - `fastctx status` – check configuration and health - `fastctx jobs` – list running jobs - `fastctx jobs kill <job_id>` – stop a job - `fastctx unapply --yes` – remove FastCtx data - `fastctx lang <code>` – change TUI language These commands work without opening the interactive terminal.
Featured in Videos
YouTube tutorials and walkthroughs for fastctx
Alternatives
Similar projects ranked by category, topics, and text overlap.
