ReckonKnow exactly where your AI spend is going.
Per-developer attribution for Anthropic, OpenAI, and Copilot. Anomaly alerts in Slack. No proxy required.
Free for up to 3 developers · no credit card · 5-minute setup
How it works
Connect provider keys
Each developer adds their own Anthropic, OpenAI, or Copilot key. No proxy, no SDK, no code changes.
We poll usage hourly
Reckon reads the providers' own usage APIs and attributes every dollar to a developer, model, and day.
Get digests + alerts
A daily Slack digest, weekly recap, and same-day anomaly alerts when someone's spend spikes.
Your spend, where your team already looks.
Install the Reckon Slack app and a daily digest posts to your channel, anomaly alerts fire the same day with one-click acknowledge, and anyone can pull numbers with /spend. Critical anomalies also file a Linear issue automatically.
📊 AI spend yesterday: $219.65 ▲ 12%
Spent $412.40 yesterday on claude-opus-4 — well above the mean + 3σ threshold for this developer.
Daily & weekly digests — totals, top spenders, and open anomalies posted to your channel.
Anomaly alerts — severity-colored, with Acknowledge and Investigate right in Slack.
/spend slash command — today, yesterday, this week, or a specific developer, on demand.
Linear integration — every critical anomaly becomes an Urgent issue, auto-closed on acknowledge.
One thing, done well.
Per-developer attribution
See exactly who spends what across every provider — joined to real people, not anonymous keys.
Anomaly alerts in Slack
When spend spikes 4× normal, you hear about it the same day — not when the invoice lands.
Slack + Linear built in
Daily digests to a channel, and a Linear issue filed automatically on every critical anomaly.
Run-rate forecasting
Project the month from the trend so a runaway agent never becomes a surprise five-figure line item.
Built to be the safest vendor you onboard.
Read-only by design
We never sit in your request path. No proxy, no TLS termination, no latency.
We never see content
We poll usage APIs for token counts and cost. Prompts and responses never reach us.
Keys encrypted with KMS
Envelope encryption per key; decrypted only inside the ingestion worker, never the web app.
Frequently asked
Do you see our prompts or responses?
No. We poll the providers' usage APIs and read what they already report. We never sit in your request path and never see content.
How does per-developer attribution work?
Each developer uses their own API key, tagged with their identity in Reckon. Provider usage APIs report at the key level, so we join usage straight through to the person — no proxy required.
Which providers do you support?
Anthropic, OpenAI, and GitHub Copilot in v1. More as customers ask.
Will this slow down our AI calls?
No. We are never in the request path, so there is zero added latency.