MindWeave v1.0 is here, check it out. #2
Mindweavecli
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Today I'm releasing MindWeave v1.0, the first public build.
It's a terminal-native coding agent that runs directly in your repo reading files, making edits, running shell commands, and verifying its own work. Everything stays on your machine. You bring your own API key, and your code never touches any server I control. Zero telemetry, no third-party backend.
I built this because too many agent tools waste half your context window on bloated prompt scaffolding before the model even reads a line of code. MindWeave keeps the system prompts thin so the model actually has room to reason through complex refactors.
What's in v1.0
The initial build includes a terminal loop with streaming UI, multi-file editing, ripgrep search, and background shell execution. Under the hood, it uses Tree-sitter and LSP integrations to give the agent full-codebase awareness without burning tokens on static structure. It also features session memory compaction so long runs don't degrade performance, plus project-level guardrails to set standing rules or block restricted paths.
For now it is set to be simple and will be be upgraded and updated overtime with new features and drivers help.
Right now, it ships with a dedicated DeepSeek driver.
Platform Support & Roadmap
v1.0 was built and tested primarily on Windows. Since the core is pure cross-platform Node, macOS and Linux support are up next if you test it on either platform and run into path or shell quirks, opening an issue is super helpful.
The architecture is built to be model-adaptive: every provider gets its own isolated driver, meaning only the code for the active model is ever loaded. Drivers for Anthropic, OpenAI, Qwen, and Ollama are coming next.
Looking for Contributors
If you know a specific model inside out, feel free to grab an open issue and take lead on its driver. AI-assisted PRs are completely fine just test your code, know how it works, and make sure it builds cleanly.
It's open source under Apache 2.0. Clone it, break it, and let me know what needs fixing.
⭐ GitHub: https://github.com/Nimannns/Mindweave
All reactions