Get a floating animated companion beside your terminal that reports what changed in your repo while you code.
Ask questions about your coding session and get grounded answers from an indexed knowledge graph of your repo, not guesses.
Wire Toshi into Claude Desktop, Cline, or any MCP client so those tools can call toshi_status, toshi_ask, or toshi_mood.
Hook Toshi to auto launch whenever you start a zero coding session.
| philpof102-svg/toshi | 0xmukesh/docusaurus-tutorial | 1tsmejp/palworld-docker-wine | |
|---|---|---|---|
| Stars | 1 | 1 | 1 |
| Language | JavaScript | JavaScript | JavaScript |
| Last pushed | — | 2021-12-27 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | easy | easy | moderate |
| Complexity | 2/5 | 2/5 | 4/5 |
| Audience | developer | developer | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Full grounded answers require indexing your repo with codebase-memory-mcp and optionally an API key for an LLM provider.
Toshi is a small animated companion that sits beside your terminal while you code, similar in spirit to the old Microsoft Clippy assistant but built around a modern AI workflow. It shows up as a floating, always on top window with the open source Toshi mascot and can talk to you about what is happening in your coding session, such as what changed or why a test is failing. The project is designed to work alongside a coding runtime called gitlawb/zero. Toshi itself is packaged as a plugin for that runtime plus a standalone MCP server, so other tools that support the Model Context Protocol, such as Claude Desktop or Cline, can also connect to it. To keep answers grounded and cheap to run, Toshi relies on a separate tool called codebase-memory-mcp, which builds a persistent knowledge graph of your repository so Toshi does not need to reread every file to answer a question. Until a repository is indexed, Toshi says plainly that it cannot read your session yet instead of making something up. Talking and chat features are optional and require connecting Toshi to a language model through any OpenAI compatible endpoint. It can auto detect API keys for providers like OpenRouter, x.ai, Groq, or OpenAI, and the model choice can be changed and saved with a simple command. Without a configured provider, Toshi still runs in a demo mode where the mascot is animated but honest about its limits rather than inventing an answer. The same codebase runs as a command line tool on Windows, macOS, and Linux, installable directly from GitHub without a package registry. Once installed, typing toshi in any terminal floats the companion, and typing it again in another terminal connects that repository to the same running instance. A setup command can also hook Toshi to launch automatically whenever the zero runtime starts. The Toshi mascot itself is used under the GPL-3.0 license, which also covers this project.
Toshi is a floating animated terminal companion that watches your coding session and answers grounded questions about your repo through an MCP server.
Mainly JavaScript. The stack also includes JavaScript, Node.js, Electron.
Copyleft license, you can use and modify the code, but derivative works must also be released under GPL-3.0.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.