Pyckle plugs deep codebase context into the editor you already use — via MCP. Keep your tools. Keep your model. Pay half what walled-garden platforms charge.
The tools you already use are great. The context layer underneath them shouldn't cost twice as much or lock you into one AI provider.
Pyckle is a context layer, not a walled garden. Switch between Claude, GPT, Gemini, or a local model without re-indexing or reconfiguring anything.
Embeddings are generated locally using PyckLM — your source code never leaves your machine. No compliance risk. No proprietary training on your codebase.
Pyckle Pro is $10/month. Transparent pricing, no "Contact Sales" wall, no per-seat enterprise contracts. Start free. Upgrade when you need to.
AI-assisted coding tools shouldn't require you to abandon your workflow to use them.
| The Old Way | Pyckle | |
|---|---|---|
| Monthly price | $20+/mo | $10/mo |
| Model lock-in | One provider only | Any model |
| Editor support | Proprietary IDE required | Any MCP-compatible editor |
| Code privacy | Cloud upload required | Local-first, code stays on-device |
| Personalized embeddings | Generic model | PyckLM — trained on your patterns |
| Cross-project context | Single repo only | Unified workspace index |
| Setup time | Platform migration required | 2 minutes, no migration |
No platform migration. No dependency nightmare. Your editor just gets smarter.
One command. No Homebrew chains, no manual config files.
PyckLM runs locally and builds a semantic index of your project. One repo or ten — it handles both.
Add Pyckle as an MCP server in your editor's config. Works with Claude Code, Windsurf, Continue, VS Code, Zed, JetBrains, and more.
Generic embedding models retrieve code. They find the file that contains a keyword. That's search.
PyckLM retrieves your code — trained on your naming conventions, your architectural patterns, your team's implicit relationships between modules. The longer you use it, the better it knows your codebase.
That specificity compounds. A context layer that improves with your codebase is a moat against any tool that resets when you switch providers.
ONNX runtime, no GPU required. Inference happens locally before any query leaves your device.
Index multiple repos into a unified workspace. Your AI sees the full picture — not just the file you have open.
Switch from Claude Code to Windsurf mid-session without losing context. The index follows you, not the tool.
PyckLM is continuously fine-tuned. Updates deploy automatically — your context gets better without any action on your part.
Start free. Upgrade when your codebase scales.
No account required to start.
Runs entirely on your machine
Or $100/year — two months free.
No credit card required. Cancel anytime.