# Relay

> Open-source Python LLM gateway by Ai5labs. One interface to every major LLM provider, with first-class MCP, streaming, tool calls, Pydantic structured output, OpenTelemetry, cost tracking, and PII redaction. Apache-2.0.

Install: `pip install ai5labs-relay` (Python 3.10+). Import name: `relay`.

## Docs

- [Agent guide](https://relay.ai5labs.com/agents.md): Self-contained install + usage guide for coding agents — minimal example, streaming, tool calls, Pydantic output, MCP, cost tracking, and the verified capability matrix for v0.1.
- [README](https://github.com/ai5labs/relay-llm#readme): Project overview, supported providers, alpha scope.
- [Benchmark methodology](https://github.com/ai5labs/relay-llm/blob/main/BENCHMARKS.md): How the cold-start and TTFT comparisons against LiteLLM were run (3 runs, honest aggregate).
- [Changelog](https://github.com/ai5labs/relay-llm/blob/main/CHANGELOG.md): v0.1 release notes.
- [Catalog (JSON)](https://raw.githubusercontent.com/ai5labs/relay-llm/main/src/relay/catalog/data/models.json): 436 models with prices, context windows, capabilities, and benchmark scores.

## Catalog browser

- [All models](https://relay.ai5labs.com/models): Searchable, filterable, sortable view of the full catalog.
- [Compare](https://relay.ai5labs.com/models/compare): Side-by-side comparison of up to 4 models.

## Support

- [Issues](https://github.com/ai5labs/relay-llm/issues)
- [Sponsor](https://github.com/sponsors/ai5labs)
- Email: engineering@ai5labs.com

## Optional

- [PyPI](https://pypi.org/project/ai5labs-relay/)
- [Privacy](https://relay.ai5labs.com/privacy)
- [Terms](https://relay.ai5labs.com/terms)