Does Bee work with Cursor or Continue?

Yes — both accept custom OpenAI-compatible endpoints. Set Bee's base URL and API key in the model settings and select a Bee model alias.

Is there a native VS Code extension?

Yes — Bee ships a VS Code extension with inline chat and completions, though any editor with OpenAI-compatible settings works without it.

Can it handle my whole repository?

256K native context covers large working sets, and the 1M-token long-context mode on Comb and Buzz handles repository-scale requests in one call.

Will my code be used to train models?

No — customer data, including code, is not used to train the Bee base model.

What does it cost to try?

Nothing — Bee Cell is free with no card required, and API Free includes 500K tokens per month for integration testing.

A coding assistant that plugs into the editor you already use

Point your editor's OpenAI-compatible settings at Bee and you have completion, chat, refactoring, and test generation — with a free tier to prove it out first.

Setup is one config block

In Cursor, Windsurf, Continue.dev, Zed, JetBrains AI Assistant, or Aider, set the base URL to https://bee.heossi.com/bee, paste a Bee API key, and pick a model alias like bee-comb. VS Code users can install the native Bee extension for inline chat on top.

Because Bee implements the Chat Completions protocol, there is no adapter layer to maintain — the editor thinks it's talking to any standard endpoint.

What it's good at day to day

→Explaining unfamiliar code and generating focused refactors — programming is an active first-class Bee domain.
→Writing tests against the actual code you paste or load, not a paraphrase of it.
→Reasoning across large slices of a repository with 256K native context, or the 1M-token long-context mode on Comb and Buzz for repo-scale questions.
→Reading screenshots of stack traces, UI bugs, or diagrams via native vision input on Comb, Buzz, and Hive.
→Returning structured JSON for tooling — lint summaries, codemod plans — with schema-guided output.

Also in your agent tools via MCP

The hosted Bee MCP server exposes chat, code, security-review, and research tools to MCP clients — Claude Desktop, Claude Code, Cursor, VS Code, Zed, Windsurf, OpenCode — with a one-line uvx install and a Bee API key. Your assistant setup and your agent setup share one account and one bill.

Cost control that fits individuals and teams

Start free on Bee Cell (128K context, no card). Individuals move to Brood at $19/mo for 256K context and more headroom; teams pool tokens on Comb ($79/mo, 3 seats). API users pay published per-token rates from $0.15/1M input on Cell.

Customer data is not used to train the Bee base model. Prompts, documents, and outputs stay yours; processing is limited to serving, securing, and operating the service.

Frequently asked questions

Does Bee work with Cursor or Continue?: Yes — both accept custom OpenAI-compatible endpoints. Set Bee's base URL and API key in the model settings and select a Bee model alias.
Is there a native VS Code extension?: Yes — Bee ships a VS Code extension with inline chat and completions, though any editor with OpenAI-compatible settings works without it.
Can it handle my whole repository?: 256K native context covers large working sets, and the 1M-token long-context mode on Comb and Buzz handles repository-scale requests in one call.
Will my code be used to train models?: No — customer data, including code, is not used to train the Bee base model.
What does it cost to try?: Nothing — Bee Cell is free with no card required, and API Free includes 500K tokens per month for integration testing.

AI infrastructure engineers don't have to fight OpenAI-compatible, down to the wire format Agents need structure. Bee returns it.

Start on the free tier

Bee Cell is free — no card. Scale to paid tiers, the API, or sovereign deployment when you are ready.

Get started free See pricing