Developers
Configure your client for Prizmal
Request access to Prizmal Service, then configure a supported client. Use one Prizmal key and model prizmal-auto. Prizmal provisions provider access; separate provider keys are not required.
Harness support matrix · connection paths for nine harnesses and two SDKs
Request access
Onboarding is manual. Tell us about your client and requirements so we can confirm supported features.
Configure the supported client
Use the examples below. Check your client's docs for current configuration keys and verify streaming, tools, and input support for your workflow.
Inspect a first run
Send a small request and inspect the response and trace before using the client in your workflow.
Codex compatibility pending
Codex requires the Responses API. Prizmal support is not yet documented; contact us before attempting setup.
Endpoints
Quick start
1# ~/.zshrc or ~/.bashrc2export ANTHROPIC_BASE_URL=https://api.prizmal.ai/anthropic3export ANTHROPIC_AUTH_TOKEN=$PRIZMAL_API_KEY4export ANTHROPIC_MODEL=prizmal-autoClaude Code appends /v1/messages to the base URL. Anthropic does not support routing Claude Code to non-Claude models; confirm Prizmal model and routing scope before deployment.
PRIZMAL_API_KEY is a placeholder, not a real key. For current API coverage and onboarding details, contact managed.inferences@prizmal.ai.
Choose your harness. Know the connection.
Harness support matrix
This matrix documents connection paths. Confirm your client version, streaming, tool calls, context limits, and required input types during onboarding.
- Setup example
- : a documented client configuration path.
- Gateway example
- : connects through the client's gateway setting, with vendor scope limits.
- SDK example
- : direct API code for the listed API only.
- Compatibility pending
- : no documented Prizmal path yet. Contact us.
Documentation reviewed 17 October 2026.
| Client | API / connection | Configuration status | Scope and limitations | Setup and docs |
|---|---|---|---|---|
| Claude Code | Anthropic Messages via LLM gatewayhttps://api.prizmal.ai/anthropic | Gateway example | Anthropic does not support routing Claude Code to non-Claude models. Confirm Prizmal model and routing scope before deployment. | |
| Cursor | OpenAI Chat Completionshttps://api.prizmal.ai/v1 | Setup example | Your own key covers chat models only. Tab and Composer are outside this setup. | |
| GitHub Copilot (VS Code Chat) | Custom endpoint, Chat Completionshttps://api.prizmal.ai/v1/chat/completions | Setup example | VS Code Chat custom endpoint only, not every Copilot product or the CLI. Agent use needs tool support. Inline completions and embeddings are outside this setup. | |
| Windsurf | Not documented | Compatibility pending | Current vendor docs do not describe a custom OpenAI-compatible endpoint. Confirm your version and workflow with us. | |
| Codex | Responses API required | Compatibility pending | Codex custom providers use the Responses API. Prizmal Responses support is not yet documented. | |
| Cline | OpenAI Compatible providerhttps://api.prizmal.ai/v1 | Setup example | Confirm model context, output limits and tool use during onboarding. | |
| opencode | Custom provider, @ai-sdk/openai-compatiblehttps://api.prizmal.ai/v1 | Setup example | The top-level model setting selects the Prizmal provider by default. | |
| Aider | OpenAI-compatiblehttps://api.prizmal.ai/v1 | Setup example | Aider has no built-in context or cost metadata for prizmal-auto. Confirm those values during onboarding. | |
| Pi | openai-completionshttps://api.prizmal.ai/v1 | Setup example | Pi reads $PRIZMAL_API_KEY from the environment. Select the model with /model. |
Claude Code
Gateway exampleAnthropic Messages via LLM gateway · https://api.prizmal.ai/anthropic
Anthropic does not support routing Claude Code to non-Claude models. Confirm Prizmal model and routing scope before deployment.
Cursor
Setup exampleOpenAI Chat Completions · https://api.prizmal.ai/v1
Your own key covers chat models only. Tab and Composer are outside this setup.
GitHub Copilot (VS Code Chat)
Setup exampleCustom endpoint, Chat Completions · https://api.prizmal.ai/v1/chat/completions
VS Code Chat custom endpoint only, not every Copilot product or the CLI. Agent use needs tool support. Inline completions and embeddings are outside this setup.
Windsurf
Compatibility pendingNot documented
Current vendor docs do not describe a custom OpenAI-compatible endpoint. Confirm your version and workflow with us.
Codex
Compatibility pendingResponses API required
Codex custom providers use the Responses API. Prizmal Responses support is not yet documented.
Cline
Setup exampleOpenAI Compatible provider · https://api.prizmal.ai/v1
Confirm model context, output limits and tool use during onboarding.
opencode
Setup exampleCustom provider, @ai-sdk/openai-compatible · https://api.prizmal.ai/v1
The top-level model setting selects the Prizmal provider by default.
Aider
Setup exampleOpenAI-compatible · https://api.prizmal.ai/v1
Aider has no built-in context or cost metadata for prizmal-auto. Confirm those values during onboarding.
Pi
Setup exampleopenai-completions · https://api.prizmal.ai/v1
Pi reads $PRIZMAL_API_KEY from the environment. Select the model with /model.
SDKs
Direct API examples. Only the listed API is documented for each SDK.
| Client | API / connection | Configuration status | Scope and limitations | Setup and docs |
|---|---|---|---|---|
| OpenAI SDK | Chat Completionshttps://api.prizmal.ai/v1 | SDK example | Chat Completions only. Other SDK APIs, including Responses, are not documented. | |
| Anthropic SDK | Messageshttps://api.prizmal.ai/anthropic | SDK example | Messages API only. Other SDK APIs are not documented. |
OpenAI SDK
SDK exampleChat Completions · https://api.prizmal.ai/v1
Chat Completions only. Other SDK APIs, including Responses, are not documented.
Anthropic SDK
SDK exampleMessages · https://api.prizmal.ai/anthropic
Messages API only. Other SDK APIs are not documented.
Client setup walkthrough
One key, one base URL, first routed request.