Skip to main content

Connect your coding tools & apps

Almost any tool that can talk to OpenAI or Anthropic can talk to Tokligence instead — you only change where it sends requests and which key it uses. This page is the hub: pick your tool below and copy the exact setting.

OpenAI-compatible base URL:  https://llm-api.tokligence.ai/v1
Anthropic Messages host:     https://llm-api.tokligence.ai
Auth:                        your Tokligence API key (tk-...)
Model:                       a real id from the catalogue — see /models
Use a real model id

Wherever a step says <your-model-id>, put an actual model from the catalogue at llm.tokligence.ai/models (for example deepseek-v4-flash). Use one that supports the format you're calling — e.g. a model that speaks Anthropic Messages for Claude Code, or Responses for Codex.

Three ways to point a tool at Tokligence​

Every integration below is really one of these three patterns. If your tool isn't listed by name, find which pattern it fits.

(a) In-app base-URL field​

Most GUI apps and editor extensions have a settings field named something like "OpenAI Compatible", "Custom API base URL", "API Host", or "Base URL". Paste https://llm-api.tokligence.ai/v1, paste your key, add a model id. This is the best route if you don't use a terminal.

GUI apps ignore environment variables

Desktop chat clients and VS Code / editor extensions do not read the OPENAI_BASE_URL environment variable — they only use their own in-app field. Set the base URL inside the app.

(b) OPENAI_BASE_URL environment variable​

Apps and scripts built on the official OpenAI SDK, and CLI tools that honor it, reroute when you set OPENAI_BASE_URL. Anthropic-SDK tools use ANTHROPIC_BASE_URL instead.

Use OPENAI_BASE_URL — the older OPENAI_API_BASE name is no longer honored.

export OPENAI_BASE_URL="https://llm-api.tokligence.ai/v1"
export OPENAI_API_KEY="tk-..."

Add these lines to ~/.zshrc or ~/.bashrc to make them persist.

On Windows without touching the terminal: press Win+X → System → Advanced system settings → Environment Variables → New (under User variables). Add one named OPENAI_BASE_URL with value https://llm-api.tokligence.ai/v1, then add OPENAI_API_KEY the same way. Open a new app window afterwards so it picks them up.

Or set them permanently from a terminal (takes effect in new windows only):

setx OPENAI_BASE_URL "https://llm-api.tokligence.ai/v1"

(c) Local proxy (power users)​

Run a local OpenAI-compatible proxy (such as LiteLLM) that forwards to Tokligence, then point your app's base URL at the proxy — for example http://localhost:4000/v1. This is niche: it's only worth it when you want local request logging, custom routing, or to combine several backends behind one endpoint.

Dead end: the hosts file won't work​

You cannot redirect api.openai.com to Tokligence by editing your hosts file. The TLS certificate for api.openai.com won't match, so the connection fails. A tool is only redirectable if it offers a base-URL field or reads the base-URL environment variable. This is why the consumer ChatGPT app and Claude Desktop can't be pointed at Tokligence — they have no such setting.

Coding CLIs & agents​

Claude Code​

Anthropic-format agent — set three environment variables and run claude:

export ANTHROPIC_BASE_URL="https://llm-api.tokligence.ai"
export ANTHROPIC_AUTH_TOKEN="tk-..."
export ANTHROPIC_MODEL="<your-model-id>"

claude

You can also put these under env in ~/.claude/settings.json. Note the Anthropic host is the bare host — no /v1.

OpenAI Codex CLI​

Codex uses the OpenAI Responses API. Add a provider block to your user-level config at ~/.codex/config.toml (on Windows, %USERPROFILE%\.codex\config.toml) and select it:

model = "<your-model-id>"
model_provider = "tokligence"

[model_providers.tokligence]
name = "tokligence"
base_url = "https://llm-api.tokligence.ai/v1"
wire_api = "responses"
env_key = "TOKLIGENCE_API_KEY"

Then set the key that env_key names and run Codex:

export TOKLIGENCE_API_KEY="tk-..."
codex "explain this repository"

Cline / Roo Code / Kilo Code (VS Code)​

In the extension settings, set API Provider to "OpenAI Compatible", then fill in:

  • Base URL — https://llm-api.tokligence.ai/v1
  • API Key — your Tokligence key
  • Model — a model id from /models (Kilo Code can auto-fetch the list)

Aider​

Prefix the model with openai/ so Aider routes it over the OpenAI-compatible surface:

aider --model openai/<your-model-id> \
  --openai-api-base https://llm-api.tokligence.ai/v1 \
  --openai-api-key tk-...

The openai/ prefix is required — without it Aider won't use your custom base URL.

Continue (VS Code / JetBrains)​

Add a model to ~/.continue/config.yaml:

models:
  - name: Tokligence
    provider: openai
    model: <your-model-id>
    apiBase: https://llm-api.tokligence.ai/v1
    apiKey: tk-...
    roles:
      - chat
      - edit

Qwen Code​

Set three environment variables and run qwen:

export OPENAI_BASE_URL="https://llm-api.tokligence.ai/v1"
export OPENAI_API_KEY="tk-..."
export OPENAI_MODEL="<your-model-id>"

qwen

Zed​

Open Agent settings → LLM Providers → Add Provider and choose the OpenAI-compatible option. Set the API URL to https://llm-api.tokligence.ai/v1, add your model id, and paste the key when prompted. You can add the Anthropic host (https://llm-api.tokligence.ai) as a separate provider the same way.

OpenHands​

Open LLM settings and toggle Advanced. Set:

  • Custom Model — openai/<your-model-id>
  • Base URL — https://llm-api.tokligence.ai/v1
  • API Key — your Tokligence key

Goose (Block)​

Configure the OpenAI provider against the bare host — Goose adds the path itself:

export GOOSE_PROVIDER="openai"
export OPENAI_HOST="https://llm-api.tokligence.ai"
export OPENAI_BASE_PATH="v1/chat/completions"
export OPENAI_API_KEY="tk-..."
export GOOSE_MODEL="<your-model-id>"

Here OPENAI_HOST is the bare host (no /v1) and OPENAI_BASE_PATH supplies the rest.

OpenCode​

Add a provider to opencode.json using the OpenAI-compatible adapter:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "tokligence": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Tokligence",
      "options": {
        "baseURL": "https://llm-api.tokligence.ai/v1",
        "apiKey": "tk-..."
      },
      "models": {
        "<your-model-id>": {
          "name": "Tokligence model"
        }
      }
    }
  }
}

Cursor​

Open Settings → Models, enable "Override OpenAI Base URL" and set it to https://llm-api.tokligence.ai/v1, put your Tokligence key in the "OpenAI API Key" field, then + Add Model with a model id from /models.

Cursor: chat & agent only

This override applies to Cursor's chat and agent features. Tab autocomplete and inline edit stay on Cursor's own backend no matter what base URL you set — that's a Cursor limitation, not a Tokligence one.

Desktop apps (no terminal needed — Windows friendly)​

If you don't use a terminal, these GUI chat apps just need the base URL and your key pasted into a settings field. The two smoothest to set up:

Cherry Studio​

  1. Open Settings (the gear icon) → Model Services.
  2. Click + Add below the provider list and choose provider type OpenAI.
  3. Set API address to https://llm-api.tokligence.ai/v1 and paste your API key.
  4. Click + Add in the model section and enter a model id from /models, then enable it.
  5. Make sure the provider toggle (top-right) is on.

Cherry Studio also has an Anthropic provider type if you'd rather use the Anthropic host (https://llm-api.tokligence.ai).

Chatbox​

  1. Open Settings → Model Provider → Add Custom Provider.
  2. Set API Mode to "OpenAI API Compatible".
  3. API Host: https://llm-api.tokligence.ai/v1 — API Path: /chat/completions.
  4. Paste your API key and enter a model id.

Others with a base-URL field​

These all work the same way — find the OpenAI-compatible provider setting and paste https://llm-api.tokligence.ai/v1 plus your key and a model id:

  • Jan — add an OpenAI-compatible provider; the base URL must include /v1 or requests 404.
  • LobeChat — add a custom OpenAI model provider with the base URL and key.
  • NextChat — set a custom OpenAI endpoint and key in settings.
  • AnythingLLM — use the "Generic OpenAI" provider (not the plain "OpenAI" one); it needs the base URL, key, and a manually entered model id.
  • Msty — add a custom OpenAI-compatible provider with the base URL and key.
  • Open WebUI / LibreChat — self-hosted (Docker / pip); point their OpenAI base URL at Tokligence.

Still stuck?​

If your tool has an OpenAI- or Anthropic-compatible setting somewhere, it will work — the pattern is always: base URL + key + a real model id. Browse models at llm.tokligence.ai/models, and see the Quickstart for raw API examples.