AI Agent Index
ByHeather MacAvelia·Independently reviewed·Published Mar 21, 2026·Updated Nov 16, 2026
Independently verified against live vendor data on Oct 19, 2026.

Open-source, local-first AI coding agent with VS Code and JetBrains plugins, a browser UI and task-planning agent fleets. Bring your own model keys or local models. Refact Cloud retired 2026.

How we scored it

Autonomy

4/5

Integrations

3/5

Pricing clarity

5/5

Evidence

2/5

Setup

4/5

The facts

Refact is an open-source, local-first AI coding agent that runs on the developer machine with model providers or local runtimes the developer configures. It was built by Small Magellanic Cloud Ai Ltd., which retired the hosted Refact Cloud service in 2026. The April 30, 2026 shutdown announcement ended hosted accounts, managed inference, credits, subscriptions and team features, and states that the project continues as open source, local-first, bring-your-own-key and community maintained.

Development now happens in the JegernOUTT/refact repository on GitHub, maintained by the most active contributor to the original project, under the BSD-3-Clause licence. The original smallcloudai/refact repository, with about 3,500 stars, was archived on May 30, 2026 and points to the new one. Releases continued after the move, with version 8.6.4 published on September 7, 2026.

Refact ships as a standalone refact binary installed with a one-line script on macOS, Linux or Windows. A resident background process serves a browser interface, a terminal interface, an LSP endpoint and the IDE plugins for VS Code and JetBrains, which the project describes as the stable surfaces, while the background process and terminal interface are an early preview. A task planner can break a feature down as task cards and run agents in isolated git worktrees, and a Buddy assistant watches the project and suggests next steps without changing code on its own.

The agent can search and edit files, run shell commands and processes, automate a Chrome browser and delegate work to subagents. Its integrations cover shell, browser, GitHub and GitLab, databases, Docker, MCP and command-line tools. Supported providers include OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, Groq, xAI and any OpenAI-compatible endpoint, and local runtimes include Ollama, LM Studio and vLLM. Refact sends requests only to the providers, runtimes and integrations the user configures, and project state is kept in a local .refact folder.

Refact itself is free and open source, and users pay only their chosen model providers. Support, bug reports and roadmap discussion run through GitHub issues and pull requests, and there is no paid support plan or SLA. The refact.ai website still carries an enterprise page offering on-premise deployment, although the shutdown announcement says there is no paid hosted service or commercial cloud support.

Developers comparing open-source coding agents may also look at Cline, OpenCode or Aider. Teams that want a hosted agent with commercial support may compare Claude Code from $17/mo billed annually or Cursor from $8/mo.

Current state Q4 2026: Refact runs fully locally after the retirement of Refact Cloud, with development in the JegernOUTT/refact repository and version 8.6.4 released on September 7, 2026. The refact.ai website still shows a banner saying Refact Cloud is shutting down soon, while the project README says the cloud service has been retired.

Pricing

free

Segment

Business

Setup

moderate

Verified

Oct 19, 2026

Transparency

Public

Contract

Month-to-month

Data training

Not Trained

Human in loop

Optional

Refact.ai pricing

Refact.ai is free to use. Checked against the vendor's own pages on Oct 19, 2026.

Refact.ai features

AutocompleteCode generationByokOpen sourceMulti file editing

Refact.ai pros and limitations

Editorial assessment

Pros

  • ✓Free and open source under BSD-3-Clause, with users paying only the model providers they choose, including local runtimes such as Ollama, LM Studio and vLLM.
  • ✓Runs entirely on the developer machine, sending requests only to configured providers and keeping project state in a local folder that can be inspected or deleted.
  • ✓A task planner runs agents in isolated git worktrees, and the agent can use shell, browser, GitHub, GitLab, database, Docker and MCP integrations.

Limitations

  • –No commercial support or SLA since Refact Cloud was retired, and development now rests largely on a single maintainer in a personal GitHub repository.
  • –The resident background process and terminal interface are labelled an early preview, with the VS Code and JetBrains plugins described as the stable surfaces.
  • –Users must install the local engine and configure their own provider keys or local models, and the refact.ai website still shows pre-shutdown messaging.

Refact.ai technical details

Deployment
idecliself-hosted
Model architectureBring your own model (OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, Groq, xAI or local runtimes)
Autonomous rateAgents plan, edit, run commands and verify results in a loop, with confirmation gates for edits and risky commands, and the task planner can run several agents in parallel worktrees.
Integrations
GitHubVS CodeJetBrainsGitLabDocker

Alternatives to Refact.ai

Rating

3.2/ 5
★★★★★

Editorial score

How we score this →

Recognition

Transparent PricingListed 2026
Embed these on your site →

Built for

EnterpriseDevToolsOpen Source

Leave a review

Used Refact.ai? Tell other buyers how it worked for you. Takes about two minutes.

Write a review

Compare

Related Content

Is this your product?

AI systems read this listing 17 times in the last 30 days. Last checked Oct 19, 2026. AI assistants repeat what this page says, so keep it current: Featured Listings are re-checked every 14 days.

Get Featured →Claim it free

AI Agent Price & Rating Tracker

Free, every two weeks. Price changes, new agent launches, acquisitions, and rating updates across the AI agent market. Verified against live vendor data, not vendor marketing.

No spam. Unsubscribe anytime. We never share your email. Read past issues