Liho Cowork
A general-purpose agent for documents, data, code and the browser. Describe the task in plain words — the agent does it and shows every step.
- Computer
- Your server
- Liho Cloud
- Browser
Agents for working with files and code. Models through one API
Hand a task to Liho Cowork, a general-purpose agent that sorts documents, changes code, works in a browser and returns the finished result. Build software in Liho IDE: on its own core, agents see data from every tool, and their changes go through your review. Connect models to your own apps through the Liho LLM API, paid in rubles.
Cowork and IDE run on macOS, Windows and Linux. Models are paid for from your Liho LLM account balance.
Liho products
Liho IDE
Its own core: agents see data from Git, CI, Kubernetes and other plugins, and the IDE adapts to their edits. AI SDLC out of the box
Liho LLM
An API to 38 models paid in rubles. One key, Chat Completions format
Liho Server
Your own Linux server: tasks and schedules keep running when your computer is off
curl -fsSL https://dl-cowork.liho.dev/server/install.sh | sudo bashLiho Cloud
A workspace on Liho servers — no hardware of your own
How to connectMobile app Coming soon
Tasks and Decisions on your phone. For now — in the browser
Start from the task, not the product
Pick a task to see which product handles it and what happens, step by step.
Hand to an agent
Write code
Build in a model
Liho Cowork
Go through a folder of PDF receipts and build a spreadsheet with totals
- Open the receipts folder to the agent
- The agent reads the files and pulls out date, vendor and amount
- A spreadsheet with monthly totals lands in your folder
Liho Cowork
Hand off a change in a project and run the tests without opening an editor
- Open the project with write access and describe the change
- The agent edits code and runs the tests in the task terminal
- A diff in Changes, commit and push from there
Liho Cowork
Rename a hundred scans to a pattern and find duplicates
- A folder of scans and a name pattern
- Renaming and duplicate search
- Unclear files become a question in Decisions, not a guess
Liho Cowork + Liho Server
Check a website every morning and send a summary, even when the laptop is off
- Schedule: every day at 09:00
- Runs on your Liho Server without your computer
- A new chat with the summary and a notification on every device
Liho IDE
Fix a bug, review the agent’s changes and run the tests
- A task for the Liho agent in the IDE chat, files via @
- Permission for every edit, review hunk by hunk
- Tests in the testing panel, commit in the Git panel
Liho IDE
Work with Claude Code or Codex next to Git, the terminal and the debugger
- Claude Code or Codex in an IDE tab with their own sign-in
- Git, terminal and debugger in the same window
- Stop and resume the session, see the tokens
Liho IDE
Investigate a failed pipeline and prepare a fix in a merge request
- A failed GitLab or GitHub Actions pipeline
- The agent reads the log and prepares a fix at the right commit
- Branch, push and merge request — with your confirmation
Liho LLM
Add a language model call to your app or bot
- Create an API key in the dashboard
- POST /v1/chat/completions with a catalog model
- Cost in the response, spending in the request history
Liho LLM
Connect your own agent to models and pay in rubles
- base_url = https://llm.liho.dev/v1
- Models from different developers with one key
- Paid from a ruble balance, limits per key
What it looks like in the products
The plan and every step on screen
Cowork shows the task plan, the commands the agent ran, and the files it read and created. Expand a step to see the command output or the changes to a file.
How Cowork works
Review with the agent, publish only when you confirm
In Liho IDE, the agent reads a merge request, its diff and discussions, and prepares a draft review or a fix. Nothing is sent to GitLab or GitHub until you confirm.
Review and CI in the IDE
Every model request and its price
The Liho LLM dashboard shows your request history: model, key, tokens, cost and response time. It also has spending charts and CSV export.
What the dashboard offers
Cowork tasks run anywhere and are managed from anywhere
Start a task on your computer, your own server or in Liho Cloud, and manage it from any device: from Cowork on another computer or from a browser.
This computer
The agent works with the folders you open, the browser and, with your permission, the desktop.
Liho Server
Your Linux server, connected with one command. Tasks and schedules keep running even when Cowork is closed.
Liho Cloud
A workspace on Liho servers, created in Cowork settings, that works without your computer.
Any device
Open another of your computers, a server or the cloud from Cowork or cowork.liho.dev: the same full app, with changes visible everywhere at once.
Where tasks run and remote controlLiho IDE runs on your computer. Remote development in the IDE is planned.
Models from different developers with one key
Prices are in rubles per 1M tokens and already include the service fee. You are charged for the actual usage of each request.
| Model | Developer | Input | Output | Context |
|---|---|---|---|---|
| GPT-6 Sol | OpenAI | ₽24.82 | ₽49.63 | 1M |
| Claude Opus 4.8 | Anthropic | ₽33.09 | ₽165.44 | 1M |
| Gemini 3.8 Flash | ₽8.272 | ₽33.09 | 1M | |
| DeepSeek V4 Pro | DeepSeek | ₽4.963 | ₽4.963 | 1M |
| Kimi K3 | Moonshot AI | ₽45.50 | ₽45.50 | 1M |
38 models from 10 developers. Catalog data as of 10 November 2026.
Where to start
Liho LLM
Your first Liho LLM API request
Create a key, send a request with curl or the OpenAI SDK and get a streamed response.
Liho CoworkCollect data from documents into a table
Open a folder of documents to the agent, describe the fields you need and check the finished table.
Liho CoworkYour own server and scheduled tasks
Install Liho Server on a Linux server, connect it to Cowork and set up a recurring task.
FAQ
Cowork is a general-purpose agent: you describe a task and it does the work itself, with documents, data, websites or a project’s code, including tests and commits. Liho IDE is a full development environment on its own core: an editor, a debugger, Git, issues, review and CI, with agents that get data from every plugin through the core and work inside it. If you want to write and debug code yourself next to an agent, choose the IDE. If you want to hand off a task and get the result, choose Cowork.
For Liho Cowork, yes: signing in to the app and access to models go through a Liho LLM account, and requests are paid from its balance.
For Liho IDE, no. Liho LLM is one way to connect models; you can use your own keys for OpenRouter, Anthropic, OpenAI or another compatible gateway.
You pay for model requests at the prices in the Liho LLM catalog, for input and output tokens. The balance is topped up in rubles. The site does not list a separate price for the Cowork and IDE apps. See the “Pricing” page for details.
Liho Cowork: macOS (Apple Silicon and Intel), Windows x64, Linux x64 (deb and rpm). Liho IDE: macOS on Apple Silicon, Windows x64, Linux x64 (AppImage and deb). Exact files and sizes are in the download center.
In Cowork you choose which folders to open and whether they can be changed, as well as how often the agent asks before acting. By default, new tasks run without isolation and in “Auto” mode: file edits and commands run without asking. You can turn on a process sandbox (on Windows, a virtual machine) and a stricter approval mode for a task.
In Liho IDE, the Liho agent asks permission before every file edit and command, and after each turn you keep or roll back its changes hunk by hunk.
The API supports POST /v1/chat/completions, including streaming and tool calling, and GET /v1/models. For basic text requests, changing base_url and the key in the OpenAI SDK is enough. The Responses API, embeddings, images and audio are not supported.
Liho IDE and Liho Cowork are distributed under the AGPL-3.0 license, but their repositories have not been published yet. See the “About Liho” page for details.
Try Liho on your own task
Download Cowork or the IDE. To use models in Cowork you need a Liho LLM account; you can create one the first time you sign in.