Integrations
Integrations overview
Connect BlazeInfer to your code through the OpenAI and Anthropic Messages APIs, or to coding agents, agent workspaces, SDKs and frameworks.
BlazeInfer works with your own code and with the tools you already use. This section shows how to connect each one and what has been tested:
- APIs: call BlazeInfer directly with the OpenAI API format, or with the Anthropic Messages API for models that support it.
- Coding agents and agent workspaces: OpenCode, Kilo Code, Pi and Orca.
- Gateways: LiteLLM, with client headers forwarded for session affinity.
- SDKs and frameworks: the OpenAI SDKs, LangChain, LlamaIndex and the Vercel AI SDK.
Compatibility at a glance
| Tool | API format | Status | Tested with | Last verified |
|---|---|---|---|---|
| OpenCode | OpenAI Chat Completions | Supported | OpenCode 2.2.5 and 1.19.5 | 2026-11-02 |
| OpenCode desktop | OpenAI Chat Completions | Supported | OpenCode 2.x desktop app with deepseek-ai/deepseek-v4.1-flash and zai-org/glm-5.3-flash | 2026-11-04 |
| Kilo Code | OpenAI Chat Completions | Supported | Kilo CLI 7.8.18 (VS Code not tested) | 2026-11-02 |
| Pi | OpenAI Chat Completions | Supported | Pi 1.5.0 | 2026-11-02 |
| Claude Code | OpenAI Chat Completions | Supported | Through LiteLLM 1.108.4 | |
| Codex | OpenAI Chat Completions | Supported | Through LiteLLM 1.108.4 | |
| Orca | Inherited from the agent | Supported | Orca 1.4.247 with OpenCode 2.2.5 and deepseek-ai/deepseek-v4.1-flash | 2026-11-03 |
| LiteLLM | OpenAI Chat Completions | Supported | LiteLLM 1.108.4 | 2026-11-10 |
Status key:
- Supported: tested end to end against BlazeInfer.
- Beta: uses a supported API format, but only part of the setup has been tested (for example, one model or one version), or none of it yet. The Tested with and Last verified columns show what was tested.
API formats
- OpenAI Chat Completions: works with every model. Use it for every coding agent that supports it.
- Other APIs (for example the Anthropic Messages API or the OpenAI Responses API) are enabled per model. The model listing in the console shows which APIs each model supports.
Any OpenAI-compatible tool
Most tools (Cline, Roo Code, Continue, Aider, Cursor's custom endpoint, LangChain, LlamaIndex, the Vercel AI SDK, and many more) need only five values:
| Setting in the tool | Value for BlazeInfer |
|---|---|
| Provider type | "OpenAI Compatible" (or "OpenAI" with a custom base URL) |
| Base URL | https://blaze.e2enetworks.net/v1 (the tool adds /chat/completions) |
| API key | Your blz_... key. Read it from the BLAZE_API_KEY environment variable if the tool allows it |
| Model | A full model ID from console, for example deepseek-ai/deepseek-v4.1-flash or zai-org/glm-5.3-flash |
| Context window (if asked) | Copy it from the console. Do not leave the tool's default |
Pick a large-context model for agents
Coding agents send large system prompts and tool definitions. Only models with a
context of at least 64,000 tokens that return tool calls with tool_choice: "auto"
are recommended; see the
console.