Connect your tools

The address for the OpenAI SDKs and any client that asks for an “OpenAI-compatible endpoint”. Send the key as Authorization: Bearer.

https://visnea-labs.com/v1

• Chat Completions — streaming (stream: true), tool calls and image input.
• Responses — the protocol Codex runs on.
• Model list: data holds the chat models for clients, models the whole priced catalog.

curl https://visnea-labs.com/v1/chat/completions \  -H "Authorization: Bearer $VISNEA_API_KEY" \  -H "Content-Type: application/json" \  -d '{ "model": "google-gemini-3-1-pro", "stream": true,        "messages": [{ "role": "user", "content": "Hi!" }] }'

# pip install openaiimport osfrom openai import OpenAIclient = OpenAI(    base_url="https://visnea-labs.com/v1",    api_key=os.environ["VISNEA_API_KEY"],)stream = client.chat.completions.create(    model="google-gemini-3-1-pro",    messages=[{"role": "user", "content": "Hi! What can you do?"}],    stream=True,)for chunk in stream:    if chunk.choices:        print(chunk.choices[0].delta.content or "", end="", flush=True)

The address for the Anthropic SDKs, Claude Code and Claude Desktop. Send the key as x-api-key or Authorization: Bearer, and anthropic-version as usual.

https://visnea-labs.com/v1/anthropic

• Messages — streaming, tools and prompt caching (cache_control).
• count_tokens — count tokens before you send.
• The Claude models in Anthropic's list format — clients fill their model pickers from it.

curl https://visnea-labs.com/v1/anthropic/v1/messages \  -H "x-api-key: $VISNEA_API_KEY" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{ "model": "anthropic-claude-sonnet-4-6", "max_tokens": 1024,        "messages": [{ "role": "user", "content": "Hi!" }] }'

# pip install anthropicimport osimport anthropicclient = anthropic.Anthropic(    base_url="https://visnea-labs.com/v1/anthropic",    api_key=os.environ["VISNEA_API_KEY"],)message = client.messages.create(    model="anthropic-claude-sonnet-4-6",    max_tokens=1024,    messages=[{"role": "user", "content": "Hi! What can you do?"}],)print(message.content[0].text)

The protocol depends on the model: Claude answers on Anthropic Messages, OpenAI's newer models on Responses, the rest on Chat Completions. Each model's endpoints and tool support are listed in the dashboard under API → Models. Agents — Claude Code, Codex, OpenCode, Hermes, Grok CLI — need models with tools.

It's the same everywhere: the Visnea address, an sk-visnea-… key and a model id. In the dashboard, under API → Connect, these steps come pre-filled with your models.

Claude CodeTerminal agent on Anthropic Messages.Claude DesktopChat, Cowork and Code through a gateway setting.CodexCLI, app and IDE on Responses.OpenCodeOpen-source agent on Chat Completions or Messages.Grok CLITerminal agent on Chat Completions.HermesNous Research's agent, either protocol.Cherry StudioDesktop client: an OpenAI or Anthropic provider.

1. 1

   Install Claude Code

   curl -fsSL https://claude.ai/install.sh | bash# ornpm install -g @anthropic-ai/claude-code

2. 2

   Point it at Visnea

   Claude Code speaks Anthropic Messages — only the address and the key change.

   export ANTHROPIC_BASE_URL="https://visnea-labs.com/v1/anthropic"export ANTHROPIC_AUTH_TOKEN="sk-visnea-…"export ANTHROPIC_MODEL="anthropic-claude-sonnet-4-6"export ANTHROPIC_DEFAULT_HAIKU_MODEL="anthropic-claude-haiku-4-5"export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1

   ANTHROPIC_DEFAULT_HAIKU_MODEL is the fast model for background chores (titles, summaries). The last line adds Visnea's models to /model.

3. 3

   Or once, in ~/.claude/settings.json

   {  "env": {    "ANTHROPIC_BASE_URL": "https://visnea-labs.com/v1/anthropic",    "ANTHROPIC_AUTH_TOKEN": "sk-visnea-…",    "ANTHROPIC_MODEL": "anthropic-claude-sonnet-4-6",    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "anthropic-claude-haiku-4-5",    "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1"  }}

4. 4

   Run it

   claude# or one-shotclaude -p "Hi! What can you do?"

   /status shows the Visnea address. Calls show up under Usage within seconds.

1. 1

   Turn on developer mode

   Help → Troubleshooting → Enable Developer Mode. The app restarts with a Developer menu.

2. 2

   Connect Visnea as a gateway

   Developer → Configure Third-Party Inference… Under Connection set Inference provider to Gateway and fill in:

   Gateway base URL      https://visnea-labs.com/v1/anthropicGateway API key       sk-visnea-…Gateway auth scheme   Bearer

   The x-api-key scheme works too — Visnea accepts the key in either header.

3. 3

   Apply

   Apply Changes — the app relaunches. On the sign-in screen continue without an Anthropic account: Chat, Cowork and the Code tab now run through Visnea and bill your balance.

   The app pulls the Claude models from Visnea on its own — anthropic-claude-sonnet-4-6, for example.

1. 1

   Install the Codex CLI

   npm install -g @openai/codex

2. 2

   Put the key in your environment

   export VISNEA_API_KEY="sk-visnea-…"

   Windows (PowerShell): $env:VISNEA_API_KEY = "sk-visnea-…"

3. 3

   Add the provider to ~/.codex/config.toml

   model = "openai-gpt-5-5"model_provider = "visnea"[model_providers.visnea]name = "Visnea"base_url = "https://visnea-labs.com/v1"env_key = "VISNEA_API_KEY"wire_api = "responses"

4. 4

   Run it

   codex# or one-shotcodex exec "Hi! What can you do?"

5. 5

   The Codex app and the IDE extension

   Both read the same ~/.codex/config.toml — restart them after editing.

   # macOS: so the app can see the keylaunchctl setenv VISNEA_API_KEY "sk-visnea-…"

   Apps opened outside a terminal don't see variables from ~/.zshrc. On Windows add VISNEA_API_KEY under Environment Variables and restart the app.

1. 1

   Install OpenCode

   curl -fsSL https://opencode.ai/install | bash# ornpm install -g opencode-ai

2. 2

   Put the key in your environment

   export VISNEA_API_KEY="sk-visnea-…"

   Windows (PowerShell): $env:VISNEA_API_KEY = "sk-visnea-…"

3. 3

   Add the provider to opencode.json

   Project root, or globally at ~/.config/opencode/opencode.json.

   {  "$schema": "https://opencode.ai/config.json",  "provider": {    "visnea": {      "npm": "@ai-sdk/openai-compatible",      "name": "Visnea",      "options": {        "baseURL": "https://visnea-labs.com/v1",        "apiKey": "{env:VISNEA_API_KEY}"      },      "models": {        "google-gemini-3-1-pro": {          "name": "Gemini 3.1 Pro"        }      }    }  },  "model": "visnea/google-gemini-3-1-pro"}

   Add the Claude models as a second provider: @ai-sdk/anthropic with …/v1/anthropic/v1.

4. 4

   Run it

   opencode

   Switch models with /models.

1. 1

   Install Grok CLI

   curl -fsSL https://raw.githubusercontent.com/superagent-ai/grok-cli/main/install.sh | bash

2. 2

   Point it at Visnea

   Grok CLI calls OpenAI-compatible Chat Completions — any model with tools fits.

   export GROK_BASE_URL="https://visnea-labs.com/v1"export GROK_API_KEY="sk-visnea-…"export GROK_MODEL="google-gemini-3-1-pro"

   Add the lines to ~/.zshrc or ~/.bashrc to keep them.

3. 3

   Run it

   grok# or one-shotgrok -p "Hi! What can you do?"

1. 1

   Install Hermes Agent

   curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

2. 2

   Put the key in your environment

   export VISNEA_API_KEY="sk-visnea-…"

   Windows (PowerShell): $env:VISNEA_API_KEY = "sk-visnea-…"

3. 3

   Add Visnea to ~/.hermes/config.yaml

   Or interactively: hermes model → Custom endpoint, then the address, key and model.

   model:  default: google-gemini-3-1-pro  provider: custom  base_url: https://visnea-labs.com/v1  key_env: VISNEA_API_KEY

4. 4

   Run it

   hermes

   Switch models with /model.

1. 1

   Add a provider

   Settings → Model Provider → + Add: name Visnea, type OpenAI.

2. 2

   Paste the key and the address

   API Key       sk-visnea-…API Address   https://visnea-labs.com

   Cherry Studio appends /v1/chat/completions itself. For Claude, add a second provider of type Anthropic at …/v1/anthropic.

3. 3

   Add the models

   Manage → fetch the model list and press + next to the ones you need, google-gemini-3-1-pro for example. Check verifies the key.