Free
$0For one developerConnect all coding agents, rerouting strategies, input and output token compression, personal usage dashboard; prepaid credits with a 5% top-up fee and 0% BYOK commission. Official source 1 ↗
Website-checked tool profile
Edgee
Agent gateway that puts Claude Code, Codex, Cursor and Copilot on a token budget, with model rerouting, token compression and per-developer cost tracking.
Sources checked
- Pricing
- Free for one developer, Team at $30 per developer per month, Enterprise on a custom quote. Tokens Edgee provides are charged at provider prices (plus a 5% credit top-up fee on Free); with your own keys or subscriptions you pay the provider directly and Edgee takes no commission. Official source 1 ↗
- Available as
- A CLI (installed with a shell script or Homebrew) launches agents through the gateway; a web console manages the organization; apps connect through the Gateway API and SDKs for TypeScript, Python, Rust and Go. · The gateway can also be self-hosted with Docker Compose or Helm, connected or air-gapped. Official source 1 ↗ · Official source 2 ↗ Official source 1 ↗
- Primary group
- Coding & Development
What it does
Edgee is an agent gateway between your coding agents and model providers. It routes requests, controls spending, reduces redundant tokens and reports usage across a team, while developers keep their existing coding tools.
Official source 1 ↗Features
- Budget-driven routing. Set a token budget per developer, squad or organization; as it fills, requests are rerouted to open-weight models mid-task, and routing continues on the next model when a provider fails or a plan cap hits. Official source 1 ↗
- Rule-based routing strategies. A strategy maps the model an agent asks for to the model that should serve it, for an exact model, a model family or everything else, and can carry a fallback for when the requested model is unavailable or returns a 429. Official source 1 ↗
- Smart routing (beta). The gateway rates each task as low, medium, high or max complexity and sends it to the model you picked for that level; classification uses a model run under zero data retention. Official source 1 ↗
- Token compression. Tool results are trimmed at the gateway and outputs kept brief, so fewer tokens are billed for the same context; Edgee's FAQ cites a 15–20% lower token bill from compression across active customers. Official source 1 ↗ · Official source 2 ↗
- Team observability. Cost and usage are broken down per developer, repo, PR and model, with sessions, debug logs and exports in the console. Official source 1 ↗
- Works with existing agents and subscriptions. Launch Claude Code, Codex, OpenCode, Cursor, Copilot, Kimi Code, Kilo Code and others through the CLI, keeping existing Claude, Codex and Cursor subscriptions. Official source 1 ↗ · Official source 2 ↗
- Bring your own keys. Register provider keys for Anthropic, OpenAI, Gemini, Vertex AI, Bedrock, Azure OpenAI, Mistral, DeepSeek, Groq and many more, or any OpenAI-compatible endpoint; keys are encrypted at rest and masked after creation. Official source 1 ↗
- Open-weight models. GLM, Kimi and MiniMax are served for a flat monthly price as routing fallbacks. Official source 1 ↗
Pricing & free limits
Team
$30Per developer per monthAdds smart routing (beta), team and squad observability, plugins for skills, hooks and MCP servers, model selection, data destinations, GitHub per-repo and per-PR attribution, spending caps; 0% top-up fee. Official source 1 ↗
Enterprise
CustomContact salesAdds private open-weight models billed per GPU hour, SSO/SAML, a private or on-prem gateway, gateway residency in the EU, US or APAC, PII redaction and ZDR controls, and a contractual SLA; credits invoiced later. Official source 1 ↗
Billing details
On Free, credits bought from Edgee carry a 5% top-up fee ($100 in credits costs $105); on Team and Enterprise there is no top-up fee. Official source 1 ↗
Seats and the Edgee subscription are billed separately from model usage. Official source 1 ↗
Free access & limits
The Free plan is limited to one developer; adding organization members, spending caps, plugins and smart routing require Team. Official source 1 ↗
Best for & limitations
Engineering leaders putting each developer or squad on a token budget for coding agents. Official source 1 ↗
Keeping a Claude Code session running when a developer hits their plan cap by falling back to another model. Official source 1 ↗
Routing an AI application's model requests through one OpenAI-compatible API. Official source 1 ↗
- Savings figures are vendor claims. The up-to-70% figure combines routing and compression; Edgee itself says its 50% SWE-bench Lite compression result is a controlled benchmark, not a forecast. Official source 1 ↗
- Smart routing still in beta. How tasks are classified and routed may still change. Official source 1 ↗
- Traffic passes through a third party. Prompts and completions transit Edgee's gateway; they are processed in memory and not stored unless debug mode is enabled. Official source 1 ↗
- Self-hosting needs a license. Running the gateway on your own infrastructure is a licensed deployment of the proprietary gateway. Official source 1 ↗
Similar tools
Tools like Edgee
Published tools that share its categories.
ColdteaAn agentic development environment spanning code, release testing, and monitoring.
Visit 
HypervibeNative Mac app that runs Claude Code, Codex, Gemini CLI and 30 more coding agents as real terminals on one visual board.
Visit 
ChatGPTOpenAI's AI assistant for chat, search, files, images, coding with Codex and agent work across your apps.
Visit 
DeepSeekFree AI chat app with web search, DeepThink reasoning and file upload, plus a pay-as-you-go API for DeepSeek's open-source language models.
Visit 
OpenPilotOpen-source Windows desktop AI agent that edits files, runs commands and researches the web with your own OpenAI-compatible model.
Visit 
GeminiGoogle's AI assistant for chat, document summaries, Deep Research reports, writing, coding, and image, video and music generation.
Visit