Edgee
Compress tokens before your LLM calls — same context, ~50% cheaper.
Edgee is a gateway that compresses tokens before they reach LLM providers like Anthropic and OpenAI, routes to faster or cheaper models, and fails over automatically when a provider is down or a plan limit is hit. Same context, roughly half the token cost.
What is Edgee?
Edgee is an agent gateway that sits transparently between AI coding agents and LLM providers to make them cheaper, faster, and more resilient. It intercepts requests to compress tokens, route to faster or cheaper models, provide automatic failover, and give teams observability into cost and usage. It works as a transparent proxy requiring no code changes, installing in under a minute.
Key features
- Token compression across input, output, and history layers for up to 50% cost reduction
- Brevity: cuts model output narration while preserving the final code
- Tool Surface Reduction (TSR): filters irrelevant MCP tools before the model sees them
- Tool Result Trimming: cleans verbose tool-result payloads by 60-90% over long sessions
- Automatic model fallback and retry when a provider hits downtime or usage limits
- Smart routing to faster or cheaper models based on preferences
- Team cost tracking with per-repo and per-PR attribution
- Bring Your Own Keys (BYOK) and transparent proxy operation with no code changes
Who it's for
- Developers running coding agents (Claude Code, Codex, Cursor, Copilot, OpenCode) who want to cut LLM token spend
- Teams needing per-repo and per-PR cost attribution and observability across coding agents
- Users who want longer, uninterrupted coding sessions by extending effective context
- Teams needing automatic failover to stay running when a provider is down or rate-limited
When not to use it
If you don't run coding agents against hosted LLM providers, or need a fully local/offline setup, an agent gateway like Edgee adds a proxy hop with little benefit. It's aimed at coding-agent workflows rather than general application LLM traffic.
FAQ
Which coding agents and LLM providers does Edgee support?
Edgee works with coding agents including Claude Code, Codex, Cursor, GitHub Copilot, and OpenCode, and is compatible with providers such as Anthropic and OpenAI. It operates as a transparent proxy, so no code changes are required.
How much does Edgee cost?
Edgee offers a free tier. The site references a pricing page for full details, but per the marketing claims the value driver is up to 50% LLM cost reduction via token compression.
How does token compression work?
Compression works across three composable layers: output (Brevity trims narration), input (Tool Surface Reduction filters irrelevant MCP tools), and history (Tool Result Trimming cleans verbose tool payloads). Edgee describes it as semantically lossless for coding tasks.
Does Edgee handle provider failover?
Yes. Edgee provides automatic model fallback and retry logic, switching providers when one experiences downtime or when usage limits are reached, so coding sessions stay running.
Will compression break my results or hurt quality?
Edgee describes its compression as 'semantically lossless for coding tasks,' reducing narration and verbose tool payloads (cut 60-90%) while preserving the final code output.
How hard is it to install and does it require code changes?
Edgee installs in under a minute and runs as a transparent proxy, so it requires no code changes. It supports Bring Your Own Keys (BYOK) for your existing provider accounts.
Share this launch
Embed this badge
<a href="https://orangebot.ai/product/edgee" target="_blank" rel="noopener noreferrer"> <img src="https://orangebot.ai/api/badge/edgee.svg" alt="Featured on OrangeBot" width="200" height="54" /> </a>