How Jents supports every tool (and what you get)
Jents captures your AI a few different ways, depending on the tool. Here's exactly how each one connects — and what you can see and measure.
The three ways Jents captures AI
- 🧾 The bill (account import) — paste one admin key from the account that pays for the AI. Jents pulls the exact dollars, turns each API key into an agent, and imports the members as people and seat holders. Minutes, org-wide, nothing to install. Anonymous: a key isn't a person.
- 📮 Telemetry (reported) — the call goes straight to the vendor; the tool just reports what it used. Per-person usage, zero latency, nothing of ours in your request path. Cost is estimated from tokens, and on a flat-fee tool it reads $0 on purpose — the money is the seat. (Claude Cowork is the exception: it reports its own real cost.)
- 🔀 Your own gateway — already run LiteLLM? Point its logs at Jents and every call arrives with exact cost and the person who made it. Nothing re-routes; your traffic and keys never leave your environment.
Routing your traffic through a Jents-hosted gateway isn't on offer right now. Telemetry already gives you the person and the bill already gives you the exact dollars, without putting us in the live path between your engineers and your provider. If you want both on the same call today, that's your own gateway. Yours, not ours.
What you get from each tool
| Tool | How it connects | What you can measure |
|---|---|---|
| Your own agent / app / script | 📮 Telemetry — Jents Capture for Python or Node — or 🔀 your own LiteLLM | usage + tokens + estimated cost — or exact cost and the person, per call, behind your gateway |
| n8n / Zapier / Make | 🧾 Give it an API key of its own, then import the bill | exact daily cost for that key, as its own agent — no per-run detail |
| LangChain / CrewAI / frameworks | 📮 Telemetry — or 🔀 your own LiteLLM | same as your own code |
| OpenAI / Anthropic API (direct) | 🧾 Admin key | exact daily cost + tokens + models, per API key |
| Claude Code | 📮 Telemetry — a block in ~/.claude/settings.json, or one managed file pushed by your device-management tool | per-person usage + adoption + seat cost |
| Claude Cowork | 📮 Telemetry — one admin setting, org-wide | per-person usage + adoption + exact cost (Cowork reports its own) |
| OpenAI Codex | 📮 Telemetry — a block in ~/.codex/config.toml | per-person usage + adoption + seat cost |
| ChatGPT Team / Enterprise | 🧾 OpenAI Admin key | the paid seat roster |
| Claude Team / Enterprise | 🧾 Anthropic Admin key | the paid seat roster |
What you can't have today
Plainly, because a guide that hedges here costs you an afternoon:
- Cursor and GitHub Copilot — nothing. They publish no telemetry and their seat imports are switched off, so there is nothing for Jents to read. Not a thinner picture: no picture.
- AWS Bedrock, Google Vertex AI, Azure OpenAI, OpenRouter — nothing. There is no cloud or aggregator account import right now.
- Routing through Jents — no. Your own gateway is the live version of it.
Two rules that make it click
1. Dollars come from the bill or from your own gateway, never from telemetry. Telemetry reports tokens, not money. Jents can price them, but that's an estimate — and on a flat-fee tool the honest per-call price is $0, because you already bought the seat. So to get real dollars onto an agent, give it an API key of its own on a connected account, or run it behind your own gateway.
2. A paid seat nobody opens is the money you can reclaim, and it takes two lists to see it. Telemetry shows who uses a tool; the admin key shows who you pay for. The gap is the waste. That pairing works today for Claude (Anthropic Admin key + Claude Code / Cowork telemetry) and for ChatGPT Team (OpenAI Admin key). It doesn't work for Cursor or Copilot, because neither list exists.
Quick "how do I add it?"
- Your bill → open Connections, paste an Admin key for OpenAI or Anthropic. → exact dollars + your seat roster.
- A tool your team uses (Claude Code, Cowork, Codex) → Add AI source → AI your team uses. → who uses it, and what it costs.
- Something you built → Add AI source → AI you build or run → "No, or it uses its own API key". → an Activation key and a snippet for your code.
- You already run LiteLLM → Settings → Gateways. → exact cost and the person, on every call.
In one line: import the bill to know what you spend, capture telemetry to know who's using it — and if you already run a gateway, connect it to get both on the same call.