Ternilo

Put ideas in motion.
An AI assistant for your files and tools, on this computer or across connected computers.

Quickstart · Documentation · Deployment · Downloads · Feedback

English · 简体中文

Ternilo is an AI assistant that reads project files, runs tools and keeps the history of its work. Use it on your computer, or connect computers and VPS instances to a Server for remote access and collaboration.

Start on your computer

Download a desktop installer or client archive for your operating system. Open the desktop app, or install the CLI executables on PATH and run:

ternilo serve --open-browser

Choose a working folder, add your model connection under Settings → Models, and send a task. Ternilo shows the response and tool activity as work progresses. The local quickstart covers installation, the first model connection and stopping the service.

Local use needs only ternilo. The desktop app and browser connect to the same local service. Interface language follows your browser's Chinese/English preference until you save a manual choice.

Use your computers remotely

Run ternilo-server on a computer or server reachable by your other devices. Use the Docker Compose guide, or follow native Server deployment.

Server starts with a protected web setup page: copy the Initialization Key from its logs, choose SQLite or PostgreSQL, and create the administrator. Once configured, restarts use the saved database and account.

Sign in to Server and open User settings → My computers. Generate a connection command and run it on the computer that will execute tasks. Open Server from your phone or another browser, choose that computer and continue working. Files and tools remain on the execution computer; it connects outbound, so its local port does not need public exposure.

For teams, the owner enables multi-user mode and grants access to specific resources. Model access and workspace access are separate permissions. See remote access and collaboration.

Choose a guide

The documentation index includes SDKs, extensions and operations. Managed execution uses an optional Worker. The planned Work container component remains a separate design.

Capabilities

Ternilo provides workspaces and sessions, streaming responses, file and command tools, persistent terminals, background tasks, attachments, history search and export. It supports OpenAI Chat and Responses, DeepSeek Responses, native Gemini and Claude Messages, plus models authorized by Server.

Presets combine plugins and tools, including skills, plans, explicitly enabled goals, scheduled tasks, subagents, MCP, LSP and signed Rhai/WASM extensions. Task permissions control file access and execution. Server adds accounts, computer management, resource sharing, model grants, usage and audit.

See capabilities and boundaries for supported behavior, isolation limits and remaining work.

Development

Source builds require the Rust toolchain in rust-toolchain.toml, Node.js and the Linorun revision pinned in .github/linorun-revision. Follow the development guide for checkout, build and validation commands, and packaging for distributable binaries.

apps/ contains programs, crates/ shared Rust modules, web/ the interface and sdk/ client libraries. Formal documentation is in docs/en/ and docs/zh-CN/; docs/development/ contains Chinese development records.

CI and releases validate the code, native platforms and deployment, then publish client/Server archives, desktop installers and the Server image. Available downloads are listed in Releases.

License

Apache-2.0. Third-party components retain their own licenses; see third-party notices.

Friends

LINUX DO