Edgee is an Agent Gateway: the layer between AI agents (Claude Code, Codex, Copilot, OpenCode, Cursor) and LLM provider APIs (Anthropic, OpenAI, GLM). It applies three things to every request. It routes under a budget you define, it compresses the payload, and it meters every token so cost can be attributed to a developer, a repo, or a PR. Up to 70% token cost reduction.

How does budget-driven routing work?

You define a routing Strategy and assign it to a developer, a squad, or the whole organization. A Strategy ties spend thresholds to routing rules. For example, on a $100 weekly Claude budget: at 75% consumption, Opus-targeted requests reroute to Kimi K3; at 85%, Sonnet requests go to GLM 5.2; at 100%, everything routes to DeepSeek. Rerouting happens mid-task, without quality loss and without interrupting the session.

Can I use my own provider API keys?

Yes. You can use Edgee’s unified access with a single Edgee API key, or bring your own provider keys for direct billing and custom models.

What visibility do engineering leaders get?

Cost and token usage per developer, per repo, per PR, per model, and per environment. Sessions are replayable, and every token is attributed rather than pooled into a single API-key line item.

Does Edgee work with Claude Code?

Yes. Install the Edgee CLI and run `edgee launch claude` to start Claude Code through the gateway. No code changes required. Edgee works with all provider plans, including consumer subscriptions such as Claude Pro and Max, not only API credits.

Which agents does Edgee support?

Edgee supports Claude Code, Codex, Copilot, OpenCode, and Cursor. The Edgee CLI wraps your agent transparently, so routing, compression, and metering happen at the network layer.

Take back control of your token spend

Name: Edgee Agent Gateway
Author: Edgee

Edgee is an Agent Gateway. It routes, compresses tokens, and observes usage from every agent. Budget-driven routing to the right model at the right price, up to 70% cost reduction.

Up to 70%Cost reduction

Mid-taskRerouting, no quality loss

5 minutesTo onboard your team

edgee — zsh

❯

Why the Edgee Agent Gateway?

Control first, savings as the result

Edgee sits between your agents and LLM providers and applies three pillars to every request:

Route: budgets and policies, per developer or per squad

Compress: fewer tokens billed for the same context. No changes to your application code

Observe: cost and usage attributed to a person, a repo, a PR

Budget-driven routing

Set a token budget per developer, per squad, or for the whole org. As a budget fills, Edgee reroutes requests to open weight models mid-task, without quality loss and without interrupting the session. Resilience comes with it: when a provider fails or a plan cap hits, routing continues on the next model in the policy.

See how routing works

Token compression

Layer 1 (Input) tool result trimming and Layer 2 (Output) brevity. Cut tool result payloads at the gateway. Semantically lossless for coding tasks: same model output, fewer tokens billed.

Learn more

Team observability

Cost and usage per developer, per repo, per PR, per model. Engineering leaders see where the spend goes before the invoice does.

Learn more

Open weight models

GLM, Kimi, and MiniMax served at turbo speed for a flat monthly price. The models your routing policies fall back to, benchmarked for coding work.

Learn more

Bring your own keys

Use Edgee’s keys for convenience, or plug in your own provider keys for billing control and custom models.

Learn more

On-premise deployment

Run the same gateway inside your own infrastructure. Prompts, completions, and provider keys never leave your network. Docker Compose or Helm, connected or air-gapped.

Learn more

Model catalog

Frontier and open weight models across every major provider, with live pricing and context sizes. Every one of them is a valid target for a routing rule.

Learn more

How to use Edgee

Whether your team runs coding agents or ships AI apps, Edgee takes over routing, metering, and compression in about 5 minutes.

For coding agents

Route your first agent in 5 minutes

Install the Edgee CLI and connect it to your coding agent. No code changes required.

Budgets from day one: assign a routing policy to a developer, a squad, or the whole org
No code changes: works as a transparent gateway for your agent
Works with any agent: Claude Code, Codex, Copilot, OpenCode, Cursor and more

Configure your coding agent

Connect Edgee to your coding agent and start routing in about 5 minutes.

1Choose your coding agent

2Install Edgee CLI

curl -fsSL https://edgee.ai/install.sh | bash

3Start saving tokens

edgee launch claude

The vision behind Edgee

Every technological shift creates a new foundation: the web had bandwidth, the cloud had compute, and AI has tokens. In a world powered by models, intelligence has a cost: tokens flow through every interaction, decision, and response.

At Edgee, we believe intelligence should move efficiently, closer to users, intent, and action. It should be compressed, routed, and optimized so decisions happen instantly. Hear from Sacha, Edgee’s co-founder, on how AI scales by mastering how intelligence moves.

Get started

Put your agents on a budget today.

Install the CLI, launch your agents through the gateway, and assign a routing Strategy per developer, per squad, or for the whole org. Routing plus compression cut token costs by up to 70%, with no code changes.

Start free See pricing