IDE Integration
Use your Drawbridge API key inside your editor. Any tool that speaks the OpenAI-compatible or Anthropic API works — set the base URL, paste your key, and pick a model.
What every tool needs
- Base URL
https://api.drawbridge-tech.com/v1- API key
- Your key from the customer console.
- Model
smart, or a specific id likeclaude-opus-4-8(see Models).
Select "OpenAI" or "OpenAI Compatible" whenever a tool asks for a provider type. Some editors append /v1 automatically — if a request 404s, try the base URL with and without the /v1 suffix.
One key for every workflow
Every tool below runs on the same key over the same base URL. Whether you want quick chat or a fully autonomous agent, one account does both — no setting to switch.
- Chat & inline edits — conversational chat and inline edits. The model returns text and diffs that you (or your editor) apply. Great for Q&A, explanations, and Cmd/Ctrl-K inline edits.
- Autonomous agents — native tool-calling, included. Autonomous agents that edit files and run commands on their own — Cursor Agent/Composer, Cline/Roo Code "Act", Codex, and Claude Code — work out of the box.
One API key does it all. Create yours in the console. See pricing for rates.
Chat & inline edits
These connect over the OpenAI-compatible endpoint.
Cursor
Cursor lets you bring your own OpenAI-compatible key. Add Drawbridge as a custom model for the chat sidebar and inline edits.
- Open Settings → Models (or Cursor Settings → Models).
- Scroll to the OpenAI API Key section and turn it on.
- Enable "Override OpenAI Base URL" and set it to https://api.drawbridge-tech.com/v1
- Paste your Drawbridge API key as the OpenAI API key.
- Under "Model Names", add a Drawbridge model id — e.g. claude-opus-4-8 or smart — then click Verify.
- Select that model in the chat input and start a conversation.
Note: Cursor's chat sidebar and inline edits (Cmd/Ctrl-K) work out of the box. Cursor Agent/Composer — the autonomous tool-calling loop — works on the same key too (see below).
Continue (VS Code / JetBrains)
Continue is configured from a YAML/JSON file. Add Drawbridge as an OpenAI-provider model. Its chat and inline-edit features fit Drawbridge well.
- Open the Continue config (config.yaml — command palette → "Continue: Open config").
- Add a model entry pointing at Drawbridge (see below).
- Reload the Continue panel and pick the model from the selector.
models:
- name: Drawbridge (Claude)
provider: openai
model: claude-opus-4-8
apiBase: https://api.drawbridge-tech.com/v1
apiKey: your-api-keyGitHub Copilot (BYOK)
Recent Copilot Chat builds support bringing your own OpenAI-compatible model. Use it for Copilot Chat (Ask).
- Open the Copilot Chat model picker → "Manage Models…".
- Choose the OpenAI-compatible / custom provider option.
- Set the base URL to https://api.drawbridge-tech.com/v1 and paste your Drawbridge key.
- Add a model id (e.g. smart) and select it in Copilot Chat.
Note: BYOK availability varies by Copilot version and subscription. Copilot Chat (Ask) works on your key; whether Copilot's agent mode can use a BYOK model depends on your Copilot version.
Agent-first tools
These drive an autonomous tool-calling loop for hands-off file edits and command execution. Native tool-use is included on your key — no separate setup.
Cline / Roo Code (VS Code)
Cline and its fork Roo Code expose a generic OpenAI-Compatible provider. Their autonomous "Act" loop drives native tool-calls — included on your key.
- Install the Cline (or Roo Code) extension from the VS Code Marketplace.
- Open the extension's settings (gear icon in the Cline panel).
- Set API Provider to "OpenAI Compatible".
- Set Base URL to https://api.drawbridge-tech.com/v1
- Paste your Drawbridge API key as the API Key.
- Set Model ID to a Drawbridge model — e.g. claude-sonnet-4-6 — and save.
Note: Cline and Roo Code are agent-first: every turn wraps your prompt in an autonomous "Act" loop that drives tools. Native tool-use is included on your key, so the loop works out of the box — and the same key handles plain chat too.
Codex CLI
Point the Codex CLI at Drawbridge by adding a custom model provider in ~/.codex/config.toml.
- Provide your key via the DRAWBRIDGE_API_KEY env var. Temporary (current session only): run export DRAWBRIDGE_API_KEY=your-api-key. Permanent: append that same line to your shell profile (~/.zshrc on macOS, ~/.bashrc on Linux), then run source ~/.zshrc — new terminals load it automatically.
- Add the provider block below to ~/.codex/config.toml.
- Run Codex with the drawbridge provider/model.
model = "smart"
model_provider = "drawbridge"
[model_providers.drawbridge]
name = "Drawbridge"
base_url = "https://api.drawbridge-tech.com/v1"
env_key = "DRAWBRIDGE_API_KEY"
wire_api = "responses"Note: Codex connects over the Responses API (set wire_api = "responses"). Its autonomous coding loop — editing files and running commands through tool-calls — works out of the box, since native tool-use is included on your key.
Claude Code
Claude Code connects through the Anthropic Messages API. Point it at Drawbridge with two environment variables — set them in your shell or in ~/.claude/settings.json.
- Temporary (current session only): export ANTHROPIC_BASE_URL=https://api.drawbridge-tech.com (no /v1 suffix) and export ANTHROPIC_AUTH_TOKEN=your-api-key.
- Permanent: put both vars in ~/.claude/settings.json (see below) so every Claude Code session picks them up — or append the two export lines to your shell profile (~/.zshrc on macOS, ~/.bashrc on Linux).
- Run claude as usual — requests route through Drawbridge.
{
"env": {
"ANTHROPIC_BASE_URL": "https://api.drawbridge-tech.com",
"ANTHROPIC_AUTH_TOKEN": "your-api-key"
}
}Note: Claude Code is an autonomous coding agent that relies on a tool-calling loop (reading/writing files, running commands). It connects through the Anthropic Messages API, which returns native tool-calls. Native tool-use is included on your key, so Claude Code works out of the box.
Related
- Pricing — One flat per-token rate
- Quickstart — Send your first request with any SDK
- Models — Model ids, context windows, and pricing
- API Reference — Endpoints, auth, and streaming
- Building Agents — Build your own agent: which mode, the tool loop, and migration