Space Bunny in your coding agent

Space Bunny Alpha is an OpenAI-compatible chat model with a 1M-token context window and a free preview — a natural fit for coding agents and CLIs that plan and edit across a whole repository. If your agent lets you set a custom OpenAI base URL and model, you can run it on Space Bunny in about a minute.

Give your agent a 1M-context tool (MCP)

Some agents — Claude Code especially — are tied to their own model and can't be pointed at Space Bunny directly. Instead, give them a tool: the space-bunny MCP server adds an ask_space_bunny action so your agent can offload work that doesn't fit its own context — analyzing a whole repository, a long document, or hours of transcripts — to Space Bunny's 1M window, then act on the answer.

Claude Code — one command installs the tool and the skill that teaches the agent when to use it:

# Claude Code — install the plugin (bundles the tool + skill)
claude plugin marketplace add codaaiteam/spacebunny-skill
claude plugin install spacebunny@spacebunny

# get a free key at spacebunnymodel.com/get-jev, then:
export SPACE_BUNNY_API_KEY=sb_live_...

Cursor, opencode, Codex, or any MCP client — add the server to your MCP config:

{
  "mcpServers": {
    "space-bunny": {
      "command": "npx",
      "args": ["-y", "github:codaaiteam/spacebunny-mcp"],
      "env": { "SPACE_BUNNY_API_KEY": "sb_live_..." }
    }
  }
}

Repos: spacebunny-mcp · spacebunny-skill

Run your agent on Space Bunny directly

If your agent lets you set a custom OpenAI base URL + model, you can use Space Bunny as the model itself.

opencode

opencode is a terminal coding agent that accepts custom OpenAI-compatible providers. There are two ways to wire it up:

• Our endpoint directly — base URL https://spacebunnymodel.com/api/v1, key sb_live_..., model space-bunny-alpha.
• Via OpenRouter — use OpenRouter as the provider and select the model stealth/space-bunny-alpha.

Full step-by-step setup for both options lives in the opencode integration guide →

Any OpenAI-compatible agent / CLI

Most agents (Codex-style CLIs, Aider, Cline, Continue, custom SDK scripts, and more) read the standard OpenAI environment variables. Point them at our endpoint and pick the model:

# Any OpenAI-compatible agent / CLI
export OPENAI_BASE_URL="https://spacebunnymodel.com/api/v1"
export OPENAI_API_KEY="sb_live_..."
# then point the agent at the model:
export OPENAI_MODEL="space-bunny-alpha"

If your tool uses a JSON or TOML config instead of env vars, the same three values apply:

{
  "provider": "openai-compatible",
  "baseURL": "https://spacebunnymodel.com/api/v1",
  "apiKey": "sb_live_...",
  "model": "space-bunny-alpha"
}

Prefer OpenRouter? Space Bunny is available there too — swap the base URL and model:

# Via OpenRouter
export OPENAI_BASE_URL="https://openrouter.ai/api/v1"
export OPENAI_API_KEY="sk-or-..."
export OPENAI_MODEL="stealth/space-bunny-alpha"

Why Space Bunny for agents

• 1M-token context — hold an entire repository, long diffs, and tool output in a single conversation, so the agent plans with the whole codebase in view instead of a narrow window.
• Reasoning model — it thinks through multi-step edits, refactors, and debugging before it acts.
• Free preview — try it in your everyday workflow at no cost while Space Bunny Alpha is in preview.
• Fast — quick enough for the tight edit-run-fix loop that agentic coding depends on.

Space Bunny Alpha is a preview model; behavior and availability may change. Always review agent-generated changes before committing.