✦ Welcome to Letronna
Letronna is an AI agent that works on your computer. It reads and edits files, runs commands, searches the web and drives a real browser right beside the chat. It comes as a desktop app and a terminal app for macOS, Windows and Linux, both built on one fast Rust core.
Use your own model keys or local models with Ollama for free, or hosted models on your Letronna account. Letronna Cloud keeps tasks running even when your laptop is closed.
📣 This channel carries news from the team: releases, events and what's next. Turn on notifications so you don't miss one.
💬 The community is @letronnacommunity: chat, help, raids, gm streaks and showcase. Every post here has a Discuss button that takes you there.
🛡️ The team will never DM you first, ask for your password, or ask you to pay.
Letronna is in alpha, so expect rough edges, and tell us what you find.
Letronna is an AI agent that works on your computer. It reads and edits files, runs commands, searches the web and drives a real browser right beside the chat. It comes as a desktop app and a terminal app for macOS, Windows and Linux, both built on one fast Rust core.
Use your own model keys or local models with Ollama for free, or hosted models on your Letronna account. Letronna Cloud keeps tasks running even when your laptop is closed.
📣 This channel carries news from the team: releases, events and what's next. Turn on notifications so you don't miss one.
💬 The community is @letronnacommunity: chat, help, raids, gm streaks and showcase. Every post here has a Discuss button that takes you there.
🛡️ The team will never DM you first, ask for your password, or ask you to pay.
Letronna is in alpha, so expect rough edges, and tell us what you find.
🚀 Letronna 0.1.0-alpha.3 is out
The biggest update yet:
• Helpers and workflows: sub-agents that split up a big job
• Auto mode: fewer approvals when you trust the task
• Browser panel beside the chat, driven by the agent
• MCP servers over HTTP with OAuth, and plugins from git
• Zed and JetBrains over ACP, plus Slack and
• Letronna Cloud: tasks, 24/7 services and web apps
•
Update with
Letronna is in alpha, so expect rough edges and frequent updates: every release and announcement is posted here first.
The biggest update yet:
• Helpers and workflows: sub-agents that split up a big job
• Auto mode: fewer approvals when you trust the task
• Browser panel beside the chat, driven by the agent
• MCP servers over HTTP with OAuth, and plugins from git
• Zed and JetBrains over ACP, plus Slack and
letronna serve for chat apps• Letronna Cloud: tasks, 24/7 services and web apps
•
letronna doctor, /usage, /insights, and errors that say *why* a request failedUpdate with
letronna update.Letronna is in alpha, so expect rough edges and frequent updates: every release and announcement is posted here first.
⚡ Built to be fast
Letronna is one Rust core, with no browser or Electron inside. On a Mac mini M4:
• Cold start in 7.65 ms
• 9.52 MB of memory at idle
• A 200k-token context fitted in half a millisecond
And it does the work: 84.6 on Harness-Bench with DeepSeek V4 Flash and 81.2 with Claude Sonnet 4.6. On a 10-task SWE-bench Verified sample it solved all ten, using the fewest tokens of the harnesses that did.
Letronna is one Rust core, with no browser or Electron inside. On a Mac mini M4:
• Cold start in 7.65 ms
• 9.52 MB of memory at idle
• A 200k-token context fitted in half a millisecond
And it does the work: 84.6 on Harness-Bench with DeepSeek V4 Flash and 81.2 with Claude Sonnet 4.6. On a 10-task SWE-bench Verified sample it solved all ten, using the fewest tokens of the harnesses that did.
🦙 Tip: run Letronna free, fully offline
No key, no bill, nothing leaves your machine:
1. Install Ollama from https://ollama.com and start it
2. Download a model:
3. Open Letronna: it finds Ollama by itself at localhost:11434
Small local models have smaller context windows, so long chats get summarised sooner. Prefer the cloud? Your own OpenAI, Anthropic or OpenRouter keys work too, and the app stays free.
No key, no bill, nothing leaves your machine:
1. Install Ollama from https://ollama.com and start it
2. Download a model:
ollama pull <model>3. Open Letronna: it finds Ollama by itself at localhost:11434
Small local models have smaller context windows, so long chats get summarised sooner. Prefer the cloud? Your own OpenAI, Anthropic or OpenRouter keys work too, and the app stays free.
☁️ Close your laptop. The work keeps going.
Letronna Cloud runs the same agent on our servers:
• Run a task and come back to the result
• Keep a service running 24/7
• Host a web app at its own address
It comes with Pro and Team, and its compute is paid from your credit.
Letronna Cloud runs the same agent on our servers:
• Run a task and come back to the result
• Keep a service running 24/7
• Host a web app at its own address
It comes with Pro and Team, and its compute is paid from your credit.
🗺 What's next for Letronna
Now: updates that install themselves, a browser the agent can use, installers that Windows and macOS trust, and a status page
Next: download the app without an account
Later: native Android and iOS apps
No dates promised: we're in alpha, and you shape it. Vote on the roadmap, or share your ideas in the community.
Now: updates that install themselves, a browser the agent can use, installers that Windows and macOS trust, and a status page
Next: download the app without an account
Later: native Android and iOS apps
No dates promised: we're in alpha, and you shape it. Vote on the roadmap, or share your ideas in the community.
