Skip to main content
Connect coding agents, browser interfaces, SDKs, and custom applications to Hivenet Router through its supported inference APIs. The right setup depends on two things:
  • the client you are connecting
  • the API format supported by the inference backend behind Hivenet Router

Claude Code

Connect Claude Code through the Anthropic Messages API.

OpenCode

Configure Hivenet Router as an OpenAI-compatible provider in OpenCode.

Pi

Add Hivenet Router and its available models to the Pi coding agent.

Open WebUI

Use Hivenet Router as the inference endpoint for a browser-based chat interface.

Use from code

Call Hivenet Router from curl, Python, JavaScript, SDKs, and custom services.

​
Choose an integration

The client talks only to the Hivenet Router router. It does not need to know which agent, machine, or inference engine serves the request.

​
Before you start

Every integration needs:
  1. A reachable Hivenet Router router URL.
  2. A client API key when authentication is enabled.
  3. An exact model ID visible to that key.
  4. A client and backend that support the same API format.

​
Router URL

For local testing:
For a remote deployment, use a protected HTTPS address:

​
Client API key

Clients authenticate with:
Use a client API key configured through static or dynamic authentication. Do not use an administrator key from:
in an application or third-party client. See API keys for key creation, model restrictions, quotas, and expiration.

​
Model ID

List the models available to the client key:
Use one of the returned IDs exactly as shown. Model names are case-sensitive, and the response is filtered according to the key’s model access.

​
Match the API format

Hivenet Router supports several client-facing inference endpoints.
Hivenet Router does not translate between OpenAI and Anthropic request formats.A request sent to /v1/messages must reach a backend that understands the Anthropic Messages API. A backend that supports only /v1/chat/completions cannot serve that request merely because both endpoints are exposed by the router.
The individual integration guides explain which endpoint and base URL the client uses.

​
Backend features still matter

Hivenet Router routes and forwards requests. It does not add model or engine capabilities that the selected backend does not support. Compatibility may depend on:
  • streaming
  • tool and function calling
  • structured output
  • system or developer roles
  • multimodal input
  • reasoning parameters
  • model-specific request fields
This is particularly important for coding agents, which often depend on structured tool calls rather than text generation alone. A successful basic chat request does not prove that tool use will work. Test the complete client workflow after connecting it.

​
Access and quotas

Integrations use the same Hivenet Router controls as direct API requests. The router can apply:
  • API-key authentication
  • model restrictions
  • request-rate quotas
  • daily token quotas
  • routing policies
  • fallback behavior
  • audit logging
  • tenant metrics
Give separate applications or integrations separate keys when you need independent access, rotation, or investigation. For example:
Keys with the same owner share quota buckets.

​
Base URLs

Clients construct endpoint paths differently. An OpenAI-compatible client commonly expects:
and adds:
An Anthropic-compatible client may expect:
and add:
Follow the exact base URL in the relevant integration guide. A misplaced /v1 can produce incorrect paths such as:
or:
These requests return 404.
1

List models

Query /v1/models with the key the integration will use.
2

Test the endpoint directly

Send a minimal curl request using the intended OpenAI or Anthropic format.
3

Configure the client

Add the router URL, API key, and exact model ID.
4

Test streaming

Confirm that output arrives incrementally when the client uses streaming.
5

Test advanced features

For coding agents, verify tool calling and file operations with a harmless task.
6

Review the request

Find the request in audit logs or metrics and confirm its tenant, model, status, and selected agent.

​
Troubleshooting

Use the direct curl test to determine whether a problem belongs to Hivenet Router or the client configuration.

​
Integration guides

Claude Code

Configure the Anthropic Messages endpoint, authentication, model, and tool-use checks.

OpenCode

Configure a custom OpenAI-compatible provider and model.

Pi

Register a custom provider and expose Hivenet Router models in Pi.

Open WebUI

Connect a browser interface and understand shared-key behavior.

Use from code

Call chat, embeddings, and reranking endpoints from your own software.

Chat completions and messages

Review Hivenet Router’s supported LLM paths and forwarding behavior.