AlphaCode – Open‑Source Rust AI Coding Agent

What it is – AlphaCode is a terminal‑first, Rust‑based AI agent that can understand natural‑language instructions, reason about a codebase, edit files, run commands, and verify the results. It also talks to a real Firefox instance and to native desktop applications, letting you automate web‑UIs and GUI programs from the same interface.

---

Core capabilities

Capability	What you can ask it to do
Code understanding & editing	Inspect a repository, plan a fix, apply patches, add tests, run the test suite and report the diff.
Shell execution	Run arbitrary commands, batch jobs, or background tasks with safety prompts.
Web automation	Through the AlphaCode Browser Agent Firefox extension it can navigate pages, fill forms, take screenshots, download files, and validate UI flows.
Desktop automation	Uses OS accessibility APIs (UI Automation on Windows, macOS Accessibility, AT‑SPI2 on Linux) to click, type, scroll, read windows, etc.
Swarm mode	Splits a large goal into independent subtasks handled by multiple agents and then aggregates the results.
Persistent sessions & memory	Saves the conversation and project context so you can resume work after closing the terminal.
Model‑agnostic	Works out‑of‑the‑box with a free built‑in model; you can also connect Claude, OpenAI, Gemini, Copilot, Cursor, Bedrock, Azure, OpenRouter or any OpenAI‑compatible endpoint.
40+ built‑in tools	File manipulation (read, write, patch), search (agentgrep), execution (bash, batch), web (webfetch, httpflow), desktop (desktop, macos_computer_use), workflow (plan, swarm), utilities (doctor, self_improve), etc.

---

Typical workflow

1. Describe a goal – e.g. “Fix the failing test in utils.rs and add a regression test.”
2. AlphaCode analyses the repository, builds a plan, and decides which tools to call.
3. Execution – it edits files, runs cargo test, maybe opens Firefox to check a UI, or clicks a desktop app.
4. Verification – runs the relevant tests again, captures screenshots or diffs, and produces a concise report.

---

Getting started (quick‑start)

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/dragonked2/alphacode/main/scripts/install.sh | bash
# Windows (PowerShell)
irm https://raw.githubusercontent.com/dragonked2/alphacode/main/scripts/install.ps1 | iex

alphacode doctor   # sanity‑check your environment
alphacode          # launch the interactive UI

You can also run a single task without the UI: alphacode run "fix the failing test".

To enable browser automation, install the Firefox extension from the link in the README and run alphacode browser setup.

---

Model & provider configuration

alphacode login                     # start OAuth / API‑key flow
alphacode provider list             # see configured providers
alphacode model use gpt-4o          # pick a model for the session

The built‑in free lane works without any keys, making the first try frictionless.

---

Safety measures

• Destructive filesystem paths are blocked.
• Risky actions trigger an explicit permission prompt.
• Shell commands run inside a sandbox with time‑outs.
• Network calls are scanned for SSRF and credential leaks.
• Sessions are persisted; crashes can be resumed.
• The agent is instructed to make small, isolated changes, rerun tests, and present a structured report before any merge.

---

When to use AlphaCode

• Bug fixing / regression hunting – let the agent locate the cause, apply a minimal patch, and verify with tests.
• Feature addition – describe the new behaviour, and AlphaCode will scaffold code, add tests, and run the suite.
• Web‑app QA – automate login flows, form validation, or screenshot‑based checks in a real Firefox window.
• Desktop‑app scripting – automate repetitive GUI tasks (e.g., opening Calculator, performing a calculation, reading the result).
• Security reviews – combine code analysis, HTTP probing, and browser‑based testing in a reproducible workflow.
• Large projects – use swarm mode to split work (code review, security audit, documentation) among multiple agents.

---

Documentation & community

• Full docs live in the docs/ folder of the repo.
• Releases are published on GitHub with versioned binaries for Windows, macOS, and Linux.
• Issues, feature requests, and bug reports are handled via the GitHub issue tracker.
• Contributions are welcomed – see the CONTRIBUTING.md for guidelines.

---

TL;DR

AlphaCode is a Rust‑native, cross‑platform AI coding assistant that goes beyond chat‑only suggestions. It can edit code, run shells, drive a real Firefox browser, and control desktop apps, all from a terminal UI, with built‑in safety guards and support for many LLM providers.

---

All information above is taken directly from the repository’s README; no external assumptions have been added.