Skip to main content
PromptQuorum
Home/Power Local LLM/Twinny Review: Archived by Its Maker, Still Open Source
Coding Assistants

Twinny Review: Archived by Its Maker, Still Open Source

Β·8 min readΒ·By Hans Kuepper Β· Founder of PromptQuorum, multi-model AI dispatch tool Β· PromptQuorum

Twinny was a free, open-source, local-first VS Code extension for AI code completion and chat, built by Richard Macarthy β€” but its official site now displays a banner stating "TWINNY HAS BEEN ARCHIVED," pointing users to an unrelated new project. The installable extension has not been updated since August 2025, though its GitHub source code remains technically un-archived and has merged at least one outside contribution as recently as September 2, 2026.

Twinny was a free, open-source (MIT-licensed) VS Code extension for AI code completion and chat, built by Richard Macarthy and designed to run against local models via Ollama by default. Its official site now displays a banner stating the project has been archived, pointing users to a new, unrelated project. This review covers what Twinny did, its unusual current split state, and what to use instead.

Twinny Review: Archived by Its Maker, Still Open Source

Key Takeaways

  • Twinny (github.com/twinnydotdev/twinny) was a free, MIT-licensed VS Code extension for AI code completion and chat
  • Built by Richard Macarthy (GitHub handle rjmacarthy), with the repository created around August 21, 2023
  • Designed to run against local models via Ollama by default, with cloud providers (OpenAI, Anthropic, OpenRouter, DeepSeek, and others) also supported
  • Its official site, twinny.dev, now displays a banner: "TWINNY HAS BEEN ARCHIVED... WE'VE MOVED ON TO SOMETHING NEW," linking to an unrelated project called Mia (mia.run)
  • The installable VS Code Marketplace extension has not been updated since August 2025
  • The GitHub repository itself is not marked archived by GitHub, and merged an outside contributor's pull request as recently as September 2, 2026

πŸ“ In One Sentence

Twinny was a free, open-source (MIT), local-first VS Code extension for AI code completion and chat, built by Richard Macarthy starting around August 2023, whose official site now declares the project archived and points to a new, unrelated project β€” even though its GitHub repository is not marked archived and has merged outside code contributions as recently as September 2, 2026.

πŸ’¬ In Plain Terms

Twinny used to be a free VS Code extension that gave you AI-powered code autocomplete and chat, designed to work with local AI models running on your own computer. Its maker has moved on to a different, unrelated project and put an "archived" banner on Twinny's website β€” but confusingly, the code on GitHub isn't formally archived and outside contributors are still submitting fixes to it. Treat Twinny as a discontinued product you can still technically use, not an actively developed tool.

πŸ“ŒNote: This review is based on Twinny's GitHub repository, its LICENSE file, its VS Code Marketplace listing, and its official site's current banner text, independently checked for this review. It does not claim PromptQuorum has run hands-on benchmarks of Twinny.

What Was Twinny?

Twinny was a free, open-source VS Code extension providing AI code completion (fill-in-the-middle autocomplete) and a chat sidebar, designed to run against local model servers by default. It was built by Richard Macarthy, with the repository created around August 21, 2023.

  • Product type: a VS Code extension only β€” no evidence of JetBrains or other IDE support was found
  • Creator: Richard Macarthy (GitHub handle rjmacarthy), with no evidence found of a formal company behind the project beyond the "twinnydotdev" GitHub organization
  • License: MIT, confirmed via the GitHub repository's LICENSE file
  • Cost: free, with no paid tier
  • Scale: 3,633 GitHub stars and 227 forks, as of this review's check

What Did Twinny Actually Do?

Twinny added fill-in-the-middle code autocomplete and an AI chat sidebar to VS Code, designed to run against Ollama by default, with several cloud providers also supported.

  • Autocomplete: single- and multi-line fill-in-the-middle code completion
  • Chat: a sidebar for conversational assistance, plus a full-screen chat mode
  • Additional features: AI-generated git commit messages, workspace embeddings, customizable prompt templates, a diff view, and basic function calling (per its release notes)
  • Local backends: Ollama by default; llama.cpp, LM Studio, and text-generation-webui are referenced in community/config discussion, though this review could not confirm the full extent of that support from a primary documentation page
  • Cloud providers: OpenAI, Anthropic, OpenRouter, DeepSeek, Cohere, Mistral AI, Perplexity, and Groq, per the project's own README

Twinny's Confusing "Archived" Status, Explained

Twinny exists in an unusual split state: its maker has publicly declared the project archived, but its GitHub source code is not formally archived and still receives outside contributions.

  • Twinny's official site, twinny.dev, currently displays: "TWINNY HAS BEEN ARCHIVED. THANKS TO EVERYONE WHO USED IT, CONTRIBUTED TO IT, AND SUPPORTED IT ALONG THE WAY... WE'VE MOVED ON TO SOMETHING NEW," linking to a separate project, Mia (mia.run)
  • Mia's own site describes itself as "a persistent AI workspace daemon that orchestrates coding agents" β€” it does not itself claim any connection to Twinny or Richard Macarthy on its own pages; the link from Twinny to Mia is one-directional and this review could not independently confirm shared authorship from Mia's side
  • The installable VS Code Marketplace and Open VSX listings have not been updated since around August 2025
  • GitHub itself does not mark the twinnydotdev/twinny repository as archived, and it merged an outside contributor's pull request fixing an OpenRouter-related bug as recently as September 2, 2026
  • Users are still opening new feature-request issues on the repository as of early September 2026, apparently without a response yet from the original maintainer confirming or denying further development

Platform, Pricing, and Licensing

Platform

What this review found:
VS Code only. No evidence of JetBrains or other IDE support.

Cost

What this review found:
Free, with no paid tier, confirmed on the VS Code Marketplace listing.

Licensing

What this review found:
MIT, confirmed via the GitHub repository's LICENSE file.

Current status

What this review found:
Declared archived by its maker on twinny.dev; the installable extension has not been updated since around August 2025; the GitHub repository itself is not formally archived and has merged at least one recent outside contribution.

Given the maintainer's own archive announcement, do not expect security patches or further official support β€” verify current status directly on twinny.dev and the GitHub repository before relying on Twinny for anything beyond casual, low-stakes use.

Who Should Still Use Twinny?

Twinny can still make sense for a narrow set of users, but almost everyone starting fresh in 2026 should pick an actively maintained alternative.

What Twinny Is Not Good For

Twinny is not a safe choice for any new, long-term, or security-sensitive setup as of 2026, given its maintainer's own archive announcement.

  • Not for anyone who needs ongoing official support or security patches β€” its maintainer has publicly moved on to a different project
  • Not guaranteed to keep working with future VS Code releases β€” the shippable extension has not been updated since around August 2025
  • Not JetBrains-compatible β€” it was built for VS Code only
  • Not an agentic coding tool β€” it offers autocomplete and chat, with no confirmed multi-step agent mode or MCP support
  • Not something to rely on for its original maintainer's continued involvement β€” future fixes, if any, would come only from outside contributors to the still-open GitHub repository

Common Mistakes When Evaluating Twinny in 2026

Most confusion about Twinny in 2026 comes from its unusual split state β€” declared archived on its own site, while its GitHub repository remains technically open.

Competitors and Alternatives

Twinny was most often compared to Continue.dev, Cline, and Tabby β€” all local-model-friendly alternatives, though Continue.dev has also since been discontinued.

Tool
Best known for
Link
ClineOpen-source, autonomous agent extension with real local-model supportCline review
TabbySelf-hosted, open-source AI coding assistant, actively maintainedTabby review
Continue.devOpen-source autocomplete/chat extension, also discontinued (acquired by Cursor, June 2026)Continue.dev review
GitHub CopilotMicrosoft/GitHub's mainstream, actively maintained AI autocomplete and chat extensiongithub.com/features/copilot

This list reflects tools commonly compared to Twinny, not an independent PromptQuorum ranking β€” verify each tool's current maintenance status before choosing, since this category has seen several tools discontinued in 2026.

Frequently Asked Questions

Is Twinny still maintained?

No, its maker has declared it archived on twinny.dev and moved on to a different, unrelated project. The installable extension has not been updated since around August 2025.

Can I still install and use Twinny?

Yes, the last shipped version is still available on the VS Code Marketplace and Open VSX, and the source code remains on GitHub β€” but expect no further official updates or support.

Is the Twinny GitHub repository archived?

No, not formally β€” GitHub itself does not mark the repository as archived, and it has merged at least one outside contributor's pull request as recently as September 2, 2026, even though the maintainer's own site declares the project discontinued.

What is Mia, and is it a Twinny successor?

Mia (mia.run) is a separate AI workspace product that Twinny's site links to as "something new" its team built. Mia's own site does not itself confirm a connection to Twinny β€” treat this as an unconfirmed, one-directional link.

Who created Twinny?

Richard Macarthy (GitHub handle rjmacarthy), with the project's repository created around August 21, 2023.

Was Twinny free?

Yes, Twinny was free and open source (MIT license) for its entire life, with no paid tier.

Did Twinny support local models?

Yes, Twinny was designed to run against local models via Ollama by default, alongside several cloud providers including OpenAI, Anthropic, OpenRouter, and DeepSeek.

Does Twinny support JetBrains IDEs?

No, this review found no evidence of JetBrains or other IDE support β€” Twinny was built for VS Code only.

What should I use instead of Twinny?

Actively maintained local-model-friendly alternatives include Cline and Tabby. Continue.dev was also a common comparison but has itself been discontinued since June 2026.

Has PromptQuorum independently tested Twinny's claims?

This review is based on Twinny's GitHub repository, its LICENSE file, its VS Code Marketplace listing, and its official site's current banner text, rather than hands-on benchmarking by PromptQuorum.

Sources

← Back to Power Local LLM