v1.2.0 -- open source -- rust

Rate limit hit? Relay it.

Your Claude Code session -- every conversation turn, file edit, error -- handed off to another agent in under 100ms. No re-explaining. No lost context.

View on GitHub Watch demo
$ npm install -g relay-dev
Relay promo

The problem

45 minutes deep. Full context. Then it stops.

Error: Rate limit reached.
Your request has been throttled.
Please wait before making another request.

-- Your context is gone.
-- Time to re-explain everything.

The fix

One command. Full context transfer. 52ms.

$ relay handoff --to codex

  [1/3] Session captured     42ms
  [2/3] Handoff built         8ms
  [3/3] Codex launched

  Handed off to codex
  Total: 52ms
62
Tests
<100ms
Capture
5MB
Binary
16
Commands
8+
Agents

Features

Everything you need to never lose context again.

relay watch

Daemon Mode

Background monitoring. Auto-detects rate limits. Hands off without you lifting a finger.

auto-cascade

Handoff Chains

First agent fails? Tries the next one. And the next. Until something works.

relay stats

Analytics

Every handoff tracked. Success rates, agent breakdown, time saved. Local SQLite.

plugins

Custom Agents

TOML file + shell script = your own agent. No Rust. Auto-discovered.

pre-send

Secret Scanning

Catches API keys, tokens, private keys before they get sent anywhere.

scoring

Smart Compression

Relevance scoring. Recent context kept. Old noise dropped. Not dumb truncation.

estimate

Cost Before Send

Shows tokens + USD price before sending to paid APIs. Free agents marked free.

retry

Resilient APIs

Exponential backoff on 429, 500-504, network errors. Three retries default.

relay replay

Test Handoffs

Re-send any saved handoff to any agent. Compare quality. Debug issues.

Works with your stack

Codex
Claude
Aider
Gemini
Copilot
OpenCode
Ollama
OpenAI
+ Plugins
Relay workflow demo

Stop re-explaining. Start relaying.

Open source. MIT licensed. Written in Rust. Used daily.

Star on GitHub