LLM Gateway for Developers with PII/API key detection, multi-provider translation, and automatic coding agent configuration.
- Real-time detection of sensitive data in LLM requests
- 30+ API key patterns (OpenAI, AWS, GitHub, Stripe, etc.)
- Multiple modes: Monitor, Redact, or Block
- 100+ language support for PII detection
- Seamless API translation between OpenAI ↔ Anthropic ↔ Gemini
- Use any model with your preferred API format
- 9.6x faster than LiteLLM
Automatically configure popular coding assistants to use the gateway:
- Continue - Auto-configures
~/.continue/config.json - Claude Code - Auto-configures
~/.claude/settings.json - Cursor - Provides setup instructions
- Codeium - Enterprise endpoint support
- Token usage by model
- Request logging
- Cost estimation
- Install the extension from VS Code Marketplace
- Configure API Keys:
Cmd/Ctrl+Shift+P→ "Tokligence: Configure API Providers" - Start Gateway: The gateway starts automatically on VS Code launch
| Command | Description |
|---|---|
Tokligence: Start Gateway |
Start the local gateway |
Tokligence: Stop Gateway |
Stop the gateway |
Tokligence: Configure API Providers |
Set up OpenAI/Anthropic/Gemini API keys |
Tokligence: Configure Coding Agents |
Auto-configure Continue, Claude Code, etc. |
Tokligence: Set Work Mode |
Choose Auto/Passthrough/Translation mode |
Tokligence: Toggle PII Firewall |
Enable/disable PII detection |
Tokligence: Open Chat |
Open the built-in chat interface |
Tokligence: Show Status |
View gateway status and configuration |
| Setting | Default | Description |
|---|---|---|
tokligence-gateway.url |
http://localhost:8081 |
Gateway URL |
tokligence-gateway.startOnActivation |
true |
Auto-start gateway |
tokligence-gateway.workMode |
auto |
Routing mode (auto/passthrough/translation) |
tokligence-gateway.piiFirewallEnabled |
true |
Enable PII protection |
tokligence-gateway.piiFirewallMode |
redact |
PII mode (monitor/redact/enforce) |
tokligence-gateway.modelRoutes |
claude*=>anthropic,... |
Model routing rules |
- Auto: Smart routing - automatically chooses passthrough or translation
- Passthrough: Direct proxy to upstream providers only
- Translation: Protocol translation only (OpenAI↔Anthropic↔Gemini)
- Monitor: Log detected PII but allow requests through
- Redact: Automatically mask sensitive data before sending to LLM
- Enforce: Block requests containing PII
| Provider | Models | API Key Pattern |
|---|---|---|
| OpenAI | GPT-4, GPT-4o, o1 | sk-... |
| Anthropic | Claude 3.5 Sonnet, Claude 3 Opus | sk-ant-... |
| Google Gemini | Gemini 2.0 Flash, Gemini Pro | AIza... |
When the gateway is running, you can use these endpoints:
| Endpoint | Protocol | Description |
|---|---|---|
/v1/chat/completions |
OpenAI | Chat API |
/v1/responses |
OpenAI | Responses API (Codex) |
/anthropic/v1/messages |
Anthropic | Native Anthropic API |
/v1beta/models/{model}:generateContent |
Gemini | Native Gemini API |
/v1/models |
OpenAI | List available models |
/health |
- | Health check |
After installing the extension:
- Run
Tokligence: Detect Coding Agents - Select "Continue" and choose "Configure"
- Continue will automatically use the gateway for all LLM requests
Your requests now have PII protection!
Use Claude with OpenAI-compatible tools:
curl http://localhost:8081/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "claude-3-5-sonnet-20241022",
"messages": [{"role": "user", "content": "Hello!"}]
}'The gateway automatically translates to Anthropic's API format.
Apache-2.0