Edgee Fleet: Govern your coding agents
You're running a company with dozens of developers and hundreds of coding agents running 24/7. Your tech team is more productive than ever — features ship faster, PRs get reviewed overnight, tests write themselves.
But Houston, you have a problem: token costs are going through the roof, with almost no visibility into where they're coming from. Every engineer spun up their own API key. Some are running Claude Code locally; others have Codex hooked into their CI pipeline. A few have autonomous agents looping on long-running tasks. You have no idea who's consuming what, or why.
Then comes the end of the month. You open the AI billing dashboard. And you feel that familiar, gut-punch moment — the end-of-month AI bill shock. You didn't see coming, from spending you couldn't track, driven by agents you couldn't control.
This is the ungoverned coding agent problem. And it's hitting every engineering org that's serious about AI tooling.
The issue isn't that AI agents are expensive — it's that they're invisible. You can't budget what you can't see. You can't optimize what you can't attribute. And you can't manage a team of engineers whose AI usage is scattered across a dozen personal API keys with no central oversight.
It is why we're launching Edgee Fleet — and with it, the observability, budget controls, and alerting stack that makes AI agent management genuinely tractable for engineering leaders.
| Fleet turns AI coding agents from ungoverned shadow spend into managed, measurable engineering infrastructure. |
|---|
One enrollment, instant governance
Fleet works by routing each team member's coding agent through the Edgee gateway. You invite a teammate, enroll their preferred tool, and Edgee creates a dedicated API key for them — with token compression enabled automatically from day one.
Supported agents out of the box: Claude Code, Codex, OpenCode and Cursor is on its way.
After enrollment, Edgee generates ready-to-use setup instructions. The team member installs the Edgee CLI and runs edgee launch claude — routing through the gateway in under a minute.
See every member. Ranked by contribution.
The Fleet dashboard gives you a card for every enrolled team member, sorted by the metric that matters: cost, tokens, or requests. Top contributors earn rank badges. Filter by any time window from 24 hours to 90 days.
Each card shows agent type, last-active status (with a live indicator for agents used in the last hour), a sparkline of activity over the period, and the period total. The overview chart at the top aggregates everything across the org — macro view and per-member breakdown on one page.
Per-key controls, without disrupting the team
Fleet's management dialog gives you per-agent controls that don't require touching your engineers' local setup:
| Token Compression Toggle agentic compression on or off per key. Enabled by default at enrollment — adjust for any member without affecting the rest of the team. | Debug Mode Capture full request and response payloads for a specific key. When a member reports unexpected agent behavior, you can investigate without guesswork. |
|---|---|
| Key Reveal & Rotation Reveal or unenroll a key directly from the dashboard. Unenrolling permanently revokes the key — re-enrollment generates a fresh one. | Compression Savings Every enrolled agent benefits from Edgee's token compression automatically. Savings are visible per member and across the org on the overview chart. |
Observability: request-level detail behind every card
Fleet's dashboard numbers are the summary. Observability is what's underneath. Every request made through the gateway by any agent, for any team member is logged with full token counts, cost attribution, compression savings, latency, and error state.
Tags flow automatically from Fleet enrollments, so you can filter the logs view by member, environment, or agent type without any manual instrumentation. When debug mode is active on a key, the logs page surfaces those entries with a dedicated icon for instant filtering.
Budget controls and alerts: governance without babysitting
The combination of Fleet enrollment and budget alerts is where FinOps teams finally get the control they need. You can set budget thresholds at three levels:
-
Per API key (per member): Alert at 50%, 80%, or 100% of a key's credit limit. Layer all three for a graduated warning system. You'll know well before anyone hits a wall.
-
Per tag over a rolling window: Track cumulative spend for a team, feature, or environment tag over 1h, 3h, 6h, 12h, or 24h. Ideal for catching runaway agent sessions before they compound.
-
Organization remaining credits: Set a floor on your total balance so you're never caught off guard by a zero-balance service interruption.
Every alert routes to email, Slack, or both — configurable per rule, with different recipients for different alert types. The alert history table shows every trigger with its type, actual value, notification channels, and resolution status. Unresolved alerts surface as a count in the sidebar so nothing slips through.
What this means for your team
For a CTO: Scale AI coding agent usage with confidence — every agent is enrolled through a governed gateway, compressed by default, and subject to the same budget controls as your production AI infrastructure.
For VP Engineering: Fleet gives you visibility into agent adoption across your team without any workflow change for your engineers. See who's active, what it's costing, and adjust per-member configurations without touching local setups.
For FinOps: Per-member cost attribution is live from day one. No log parsing, no tagging strategy required. Set per-key budget alerts and get Slack notifications the moment anyone approaches their limit.
Fleet, Observability, Budgets, and Alerts are all available today. If your organization is already on Edgee, open Fleet in the sidebar and enroll your first team member.
| Enroll your first coding agent in under a minute Fleet is live in the Edgee Console. Invite a teammate, select their agent, and you're done. Open Fleet → |
|---|