Documentation Index

Fetch the complete documentation index at: https://to11.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

​

What is to11?

to11 is an open-source AI engineering platform that unifies LLM proxying, security, observability, and evaluation into a single product. Instead of stitching together separate tools for tracing, guardrails, routing, evals, and alerting, to11 provides one integrated platform on a shared data plane.

​

Who is to11 for?

to11 is built for AI engineers and platform engineers who:

• Ship LLM-powered applications or agents to production
• Need observability across non-deterministic LLM behavior
• Want inline security (PII detection, prompt injection blocking) without adding latency
• Use multiple LLM providers and need a unified interface
• Need to answer “why did this agent fail?” and “is our AI leaking PII?” from the same dashboard

​

Key Capabilities

​

Gateway

A Rust-based LLM proxy that sits between your application and upstream providers. It adds less than 1ms of overhead while providing:

• Cross-format routing — Send requests in OpenAI format, route to Anthropic (or vice versa). The gateway translates between wire formats automatically.
• Inline security — PII detection, blocklist filtering, and ML-based content moderation run before requests leave the gateway.
• Streaming — Zero-copy SSE passthrough when formats match; full normalization when they differ.
• W3C Trace Context — Distributed tracing propagation across multi-call and multi-agent workflows.

​

Telemetry

OpenTelemetry-native observability built on GenAI semantic conventions:

• GenAI spans — Every LLM call produces a span with model, tokens, latency, and optionally prompt/completion content.
• GenAI metrics — Histograms for operation duration, token usage, TTFT, and stream duration.
• ClickHouse storage — All telemetry lands in ClickHouse via a custom OTel Collector build with ES256 OIDC authentication.

​

Dashboard

A Next.js web application for exploring traces, reviewing LLM calls, running evals, and configuring alerts.

​

Architecture

to11 follows a four-service architecture:

1. Gateway (Rust) — LLM proxy, sync guardrails, telemetry emission
2. OTel Collector (Go) — OTLP ingestion, ClickHouse export
3. API (Effect-TS) — REST API, API key management, OIDC endpoints
4. Web (Next.js) — Dashboard, analytics

All services share ClickHouse as the telemetry data plane and PostgreSQL as the control plane.

​

Next Steps