Edgee

Agent gateway that routes and compresses tokens to cut LLM spend.

Intermediate
Screenshot of Edgee, Agent gateway that routes and compresses tokens to cut LLM spend

What is Edgee?

Edgee is an agent gateway for teams running coding agents or AI apps. It sets token budgets, reroutes to open weight models mid-task, trims and shortens payloads, and reports cost per developer, repo and PR, with BYOK and on-premise options.

Edgee is an agent gateway: it sits between coding agents or AI applications and LLM providers, then routes, compresses and meters every request. The stated goal is cost control, with the vendor claiming reductions of up to 70 percent through a combination of budget-driven routing and token compression. Rather than swapping models by hand, teams set a token budget per developer, per squad or for the whole organization, and Edgee moves traffic to cheaper open weight models as that budget fills. The switch can happen mid-task and, according to the homepage, without quality loss or an interrupted session. The routing layer doubles as a resilience layer. If a provider goes down or a plan cap is reached, requests continue on the next model allowed by the policy. Compression works in two stages: input-side trimming of tool result payloads and output-side brevity, described as semantically lossless for coding tasks. Observability attributes cost and usage to a person, a repository, a pull request or a model, so engineering leaders can inspect spend before the invoice arrives. Open weight models such as GLM, Kimi and MiniMax are served at what the site calls turbo speed for a flat monthly price, and a catalog covers frontier and open weight models across major providers with live pricing and context sizes. The product targets engineering teams running Claude Code, Codex, Copilot, Cursor, OpenCode and similar agents, plus teams shipping their own AI applications. Onboarding is a CLI install followed by launching the agent through the gateway, which the site puts at roughly five minutes with no application code changes. Keys can be Edgee's own or bring-your-own for billing control and custom models. An on-premise option runs the same gateway inside a private network via Docker Compose or Helm, connected or air-gapped, keeping prompts, completions and provider keys local. Edgee is best understood as infrastructure rather than an agent itself. Teams comparing it with per-provider dashboards or simple API proxies get policy-driven routing and compression in one place, but they also take on a gateway dependency in the request path that has to be evaluated against their own workloads.

How do you use Edgee?

  1. 1Install the Edgee CLI
    Run the install script on macOS or Linux (Homebrew is also listed) and confirm the binary is in place before connecting an agent.
    Edgee — Install the Edgee CLI
  2. 2Review available models and pricing
    Check the model catalog for frontier and open weight options with live pricing and context sizes, since every listed model can be a routing target.
    Edgee — Review available models and pricing
  3. 3Launch your coding agent through the gateway
    Start the agent with the edgee launch command, for example edgee launch claude, so its requests pass through the gateway without code changes.
  4. 4Define budgets and routing policies
    Assign a routing policy to a developer, a squad or the whole organization so requests move to fallback models as budgets fill.
    Edgee — Define budgets and routing policies
  5. 5Decide on keys and deployment
    Choose between Edgee keys and bring-your-own provider keys, and pick hosted or on-premise if prompts and keys must stay internal.
    Edgee — Decide on keys and deployment

Pros and cons

Pros

  • Budget policies can be assigned per developer, per squad or org-wide, and reroute traffic automatically as spend fillsAI
  • Works with Claude Code, Codex, Copilot, Cursor, OpenCode and other agents as a transparent gateway with no application code changesAI
  • Compression trims tool result payloads and shortens output, described as semantically lossless for coding workAI
  • Usage is attributed per person, repo, PR and model, which gives leaders visibility before invoicingAI
  • On-premise deployment via Docker Compose or Helm keeps prompts, completions and provider keys inside the networkAI

Cons

  • The headline 70 percent cost reduction is a vendor claim; actual savings depend on how much traffic can move to cheaper modelsAI
  • Mid-task model switching introduces a routing layer in the request path that teams must test against their own quality barAI
  • Compression is pitched as lossless for coding, so workloads outside coding may see less benefit or need separate validationAI
  • Open weight models are positioned for coding tasks specifically, so general or multimodal workloads may find fewer fallback optionsAI
  • Savings and observability only apply to traffic that runs through the gateway, so unmanaged agents stay invisibleAI

How much does Edgee cost?

Support

A contact page is available for reaching the team; the homepage does not describe specific support channels or service levels.

Learn more

Integrations

The gateway is designed to front coding agents including Claude Code, Codex, Copilot, Cursor and OpenCode, plus custom AI apps, and connects to major LLM providers in the model catalog.

Learn more

Features

Three pillars per request: routing under budget policies, input and output token compression, and per developer, repo, PR and model observability. Supporting pieces include a model catalog with live pricing, open weight models served for a flat monthly price, bring-your-own keys, and on-premise deployment via Docker Compose or Helm.

Learn more

Frequently asked questions about Edgee

  • How do you use Edgee?
    The walkthrough on this page covers 5 steps: 1. Install the Edgee CLI 2. Review available models and pricing 3. Launch your coding agent through the gateway 4. Define budgets and routing policies 5. Decide on keys and deployment.
  • What platforms does Edgee support?
    Edgee is available on Web App, MacOS, Windows and Linux.
  • What does Edgee integrate with?
    The gateway is designed to front coding agents including Claude Code, Codex, Copilot, Cursor and OpenCode, plus custom AI apps, and connects to major LLM providers in the model catalog.
  • What are the limitations of Edgee?
    The headline 70 percent cost reduction is a vendor claim; actual savings depend on how much traffic can move to cheaper models. Mid-task model switching introduces a routing layer in the request path that teams must test against their own quality bar. Compression is pitched as lossless for coding, so workloads outside coding may see less benefit or need separate validation.

Status

StatusActive
Views0
Outbound clicks0
Added11/3/2026

Platforms

Web AppMacOSWindowsLinux

Pricing

FreemiumPaidSubscription