Clawgate sits between your developers and AI providers, giving engineering teams complete visibility, governance, and cost control across AI coding tools like Claude Code, Codex, and opencode.
Monitor usage. Enforce budgets. Reduce token usage. Optimize model selection. Track every AI dollar by project. The savings AI promised, without the surprise bill.
The question every leader faces
Do you know what your AI coding tools are actually costing you?
Can you see who is using AI, where it's being used, and whether you're paying more than you need to?
Clawgate answers all
When you give AI to your team, you can’t see the spend until the invoice arrives. The biggest companies in the world are finding this out the hard way.
Used up its entire 2026 AI budget in four months.
After giving Claude Code to about 5,000 engineers, the cost per engineer hit $500 to $2,000 a month and blew past the budget. Their answer was blunt usage caps. Clawgate goes further: set hard limits and decide which models your team can use, so you serve lower-cost models when full power isn’t needed.
Source: Fortune →
Moved to cap employee AI usage as costs approached billions.
Staff used 73.7 trillion tokens in just 30 days, racing to top an internal leaderboard. Meta is now building usage tracking and budgets, because, in its CTO’s words, “token usage alone is not a measure of impact of any kind.”
Source: The Information →Staff can run company AI on private side projects. The company still pays for it.
Even when you trust your team, you can’t tell which project cost what. The bill is just one big number.
Claude Code, Codex, and opencode pick the model on their own. You pay for that choice, with no say in it.
Companies can save millions by giving engineers AI. But the moment you lose control of how it’s used, those savings turn into waste and surprise bills.
Clawgate is that control layer. Point Claude Code, Codex, or opencode at Clawgate with a
vsk_…
key. Every request is checked against your budget and tied to a user and a project
before it ever runs.
Daily and weekly token, session, and USD caps per user and per project. When the cap is hit, the request stops.
Pick which Claude and GPT models your team can use. You decide what runs, so you control what it costs.
Per-user and per-project cost breakdowns, transparent pass-through pricing, and monthly invoices.
AI makes your team faster either way. The real question is whether the cost climbs along with the output, or races past it.
The bill blows past the budget
The bill stays inside the budget
With Clawgate, the gains from AI productivity stay in your business, instead of leaking out as an ungoverned bill.
This is an illustrative model, so the shape matters more than any single figure. It assumes a team of about 30 developers using Claude Code, Codex, and opencode daily, with per-developer spend anchored to the $500 to $2,000 per engineer each month that companies like Uber reported above. The budget grows with output. Left ungoverned, spend drifts above it as always-on top-tier models, retries, and side projects pile up, while governed spend stays inside the cap for the same output.
These aren't numbers we invented for a slide. Clawgate started as an internal tool at Virstack LLC: we built it to keep our own engineers' AI spend under control before we offered it to anyone else. The pattern above is the one we watched play out on our own bill first.
Illustrative model. A 200-developer org at a mid-range $800 / dev / month of uncontrolled AI spend.
Compression is on by default: the model assumes it takes ~13% off the bill, and that share grows the longer a session runs — the deeper the context, the more repeated history there is to compress. Token counts fall much further than cost, because the parts that dominate the bill — output tokens and cache reads — are not compressible.
Model routing sends the work that does not need a frontier model to an open one like GLM 5.2 or DeepSeek V4, so its saving comes off the residual bill rather than the starting one. Run it on today's models alone and the same org still lands at $1.57M — $350K saved, 18% lower.
Clawgate can compress bulky tool output before it reaches the model. Old context gets squeezed, fresh work stays untouched, and accuracy holds. One switch per organization. Savings show up on your dashboard in dollars.
Accuracy unchanged
Accuracy improved
at 19% compression
at 32% compression
Measured on real agent workloads and standard public benchmarks. Recent context and model reasoning are never altered.
Full benchmarks & methodology →No new tools for your engineers to learn. It’s a one-line endpoint change.
Point Claude Code, Codex, or opencode at your Clawgate gateway and hand each developer a vsk_… key. That’s the whole setup.
Define budgets, allowed/forced models, and per-project rules with simple toggles. No code, no config files.
Track spend per user and project, get monthly invoices, and surface abuse signals before they cost you.
One command points your CLI at Clawgate — it prompts for the developer’s
vsk_… key and writes the config for you.
curl -fsSL https://console.clawgateai.com/config/claude | bashcurl -fsSL https://console.clawgateai.com/config/codex | bashcurl -fsSL https://console.clawgateai.com/config/opencode | bashClawgate Inference brings leading models from Anthropic, OpenAI, and more into Claude Code, Codex, and opencode, with no extra setup for your team. You choose which models they can use, and usage is billed at provider rates, at cost, plus one clear platform fee.
Claude
GPT
Grok
DeepSeek
Kimi
GLM
Fugu
Nemotron
New models are added as providers release them. Admins set the allowed models per team and per project, so you can serve lower-cost models whenever full power is not needed.
On the Team plan and above, connect an OpenRouter key to reach models far beyond the curated catalog above. The same per-team and per-project allow-lists, budgets, and analytics apply.
Daily & weekly token, session, and USD caps per user and per project. Spend literally cannot exceed the cap.
Pick which Claude and GPT models your team can use. You decide what runs, so you can serve lower-cost models and control the cost per request.
Scope every request to a project with the x-project-id header. Finally know which project cost what.
Real-time usage and cost dashboards, per-user breakdowns, and history retained from 7 days up to unlimited.
Privacy-safe fingerprints flag shared keys, content mismatches, and usage spikes. We never read your raw code.
AI costs passed through at cost with a clear platform fee. Cache savings passed on. Failed requests are never billed.
Squeeze bulky tool output before it reaches the model. Up to 92% fewer tokens on heavy workloads, with accuracy intact. Toggle it per organization.
Setup is a single endpoint change for your team. After that, everything is dashboard toggles: set a budget, pick the models, watch the spend. No surprises, no waste. Just the savings AI was supposed to deliver.
You pay for AI at cost, plus a clear platform fee. Budgets are hard stops, so the bill can’t run away.
Full cost transparency, metered.
$0 minimum · up to 10 users
Start freeBudgets, history, and exports.
90-day history · up to 20 users
Get StartedHow to run AI under control and know exactly what it costs.
One prompt, six top models: the cheapest run cost 26¢ and the priciest over $3 — a 12× spread for the same task. See real per-model cost and stop overpaying.
Run GLM 5.2 — an open-weight model near Opus benchmarks at a fraction of the cost — inside Claude Code and Codex with one key, dashboard model control, and automatic usage tracking.
Give your team the AI they need. Give yourself the visibility and control you need.