[00] What we do
Tokenoscopy sees every agent action and stops runaway spend. ‘Dangerously skips permissions’ never felt as safe.
npm i -g tokenoscopy && tokenoscopy initNo key to paste: init opens your browser to pair the machine.
- Works with
- Claude Code
- Codex CLI
- Anthropic SDK apps
- Cursor· next
- Gemini CLI· next
Your side · laptop · CI · prod
→ DIRECT TO ANTHROPIC / OPENAI · NO PROXY
API key = billed · seat = usage limit
↓ REDACTED EVENTS ↑ ALLOW / DENY ≤1,500ms
● Tokenoscopy Cloud
Replays every session · prices it in its lane · stops the runaway
tokenoscopy CLI
Free, on every machine that runs an agent. Records, redacts and guards.
Tokenoscopy Cloud
Free for one developer. Replays every session and stops the runaway.
[01] The CLI
One command on the machine. It never sits between your agent and the model.
| For | Every machine that runs an agent: laptops, CI runners, fleets |
|---|---|
| Runs | Inside Claude Code's own hooks, and in-process via monitor() |
| Works with | Claude Code · Codex CLI · Anthropic SDK apps · Cursor (next) · Gemini CLI (next) |
| Price | $0 |
0
proxies in your path
Records. Every prompt, model call, tool I/O and diff, priced per request and attributed to repo, branch, developer and model.
On-device
redaction
Redacts. Keys, tokens, JWTs, private keys and emails are scrubbed before upload, then again server-side.
1,500ms
guard ceiling
Guards. Answers each tool call inside the ceiling or allows it. Allows are cached 15s; deny is never cached.
CI: init --project + TOKENOSCOPY_KEY · Fleet: init --print → MDM · install docs
[02] Tokenoscopy Cloud
Where the recording lands. Replay it, price it in its lane, and stop it next time.
| For | The developer or team paying for agents |
|---|---|
| Runs | Hosted. Recording never stops on any plan |
| Works with | Claude Code sessions · production apps via monitor() |
| Price | $0 for one developer · Solo $29/mo after a 14-day trial · Team custom |
[ Services ]
Production
| Service | Status | This hour | Today (UTC) | Stop-limits |
|---|---|---|---|---|
| invoice-api | active | $18.20 | $96.40 | $18.20 / $25.00 per hour → block heads-up at 70% · 90% |
| checkout-api | active | $2.10 | $31.75 | $2.10 / $10.00 per hour → alert heads-up at 70% · 90% |
[ What fired ]
More calls, not bigger ones. requests/hour: 1,240 now vs 180 usual.
By model
By token class
[03] Where spend leaks
Three meters. Each in its own lane, each with a hard stop.
Claude Pro & Max seats
Seats feel free, so nobody sees whose loop is burning through the team's usage limits.
It stops
Seat-scoped budgets stop a runaway session before its next tool call, and a pace alert says days ahead when a cap will be reached.
How →Claude Code on API keys
An agent in CI loops overnight, and the first record of it is a four-figure invoice.
It stops
Caps per session, repo, developer or org. At the cap, the guard denies the next tool call; the status line shows the cap in the terminal before that.
How →Your apps that call Claude
A retry storm, or a prompt that quietly doubled in size, found at the end of the month.
It stops
Stop-limits per service: heads-ups at 70% and 90%, then calls refused within ~15s. No admin key.
How →[04] What stays yours
Your traffic, your machine, your call. It runs inside your agent's hooks, so it is never the reason a session stalled.
1,500ms
guard ceiling
The one hook in your critical path answers inside this or the tool call proceeds. Down, slow, or wrong, it allows.
How →
15s
allow cache
An allow is reused locally, so most tool calls never touch the network. Deny and ask are never cached.
How →
0
proxies in your path
No gateway, no proxy. Agent hooks are async, so Claude Code never waits on them. In production, the wrapper reads token counts inside your process. Your traffic never routes through us.
How →
On-device
redaction
Keys, tokens, JWTs, private keys and emails are scrubbed before upload, then again server-side.
How →
[05] Pricing
Cheaper than what it catches. Priced on seats, never on your spend.
Free
For one developer, for good
$0forever
no card · no call · one command to install
- Every session recorded and replayable
- Findings that say what to stop and what to fix
- One per-session cap that stops a runaway
- Email heads-up at 50, 80 and 100%
- 7-day replay history
Solo · most teams start here
For one developer and their apps
$29per month
14-day free trial · no card · or $290/yr
- Every session recorded and replayable
- Every budget enforced: session, repo, developer, org
- Production stop-limits on your apps' API spend
- Slack & email heads-ups before any cap
- Weekly Flight Report · 90-day history
Team
For engineering teams of any size
Customper team
priced on seats, never on your spend
- Everything in Solo, for every developer
- Who spent what, by repo, dev and model
- Seats and API keys side by side, never summed
- MDM rollout, SSO, audit export, org policies
- Priority support
Every workspace starts with a 14-day Solo trial, then drops to Free. Recording never stops on any plan; plans differ in what blocks and how long the replay detail is kept.
[06] Before you install
The questions devs ask first. Every answer links to the docs that back it.
Do you proxy my model traffic?
No. For coding agents, Tokenoscopy rides the agent's own hooks and transcripts, in Claude Code and in Codex CLI. For production apps, a wrapper inside your process reads the token counts on each response, never prompts or outputs. Either way, requests go straight to Anthropic or OpenAI with your own key. Read more →
What leaves my machine?
Session events and transcripts, redacted on your machine before upload: credential shapes, JWTs, private-key blocks, Authorization headers, connection-string passwords, secret assignments and emails in content. The server runs the same scrub again. Replays include the paths of files the agent read or edited, plus branch drift counted from local git refs. No GitHub access. Read more →
What happens if Tokenoscopy is down?
Nothing you'd notice. The budget guard has a hard 1,500ms timeout and every failure path allows the tool call: outage, network flake, rate limit, bad response. Only an explicit over-budget verdict blocks. We are never the reason an agent stopped. Read more →
Is that dollar figure money?
Depends on the lane, and we always say which. On an API key it's real money, estimated at list price. On a Claude Pro/Max or ChatGPT plan seat it's what the work would have cost at API rates. Nobody is invoiced; what runs out is your usage limit. The two are never added together. Read more →
I sign in with OAuth. Doesn't that mean subscription?
Not necessarily. Claude Code also signs into Console accounts that bill per token. We read the plan your Claude login reports instead of assuming, and a machine we can't classify is labelled unclassified rather than guessed. Read more →
Which agents are supported?
Claude Code and Codex CLI today, on laptops and in CI, plus any app that calls Claude through the TypeScript Anthropic SDK. Cursor and Gemini CLI are next. Read more →
Can it stop spend in my production app?
Yes, without an Anthropic admin key. Wrap your Anthropic client with monitor() from tokenoscopy/anthropic and give each service a stop-limit per hour or day. Slack and email get a heads-up at 70% and 90%; at the cap the wrapper refuses calls within ~15s. If Tokenoscopy is unreachable, every call goes through. Read more →
Can I roll it out without touching every laptop?
Yes. Render the hook block once with init --print and ship it through managed settings (MDM). Managed policy can't be disabled by users. CI checkouts use init --project with the key in the environment. Read more →
Is the name what I think it is?
Yes. Nobody enjoys the prep, and everyone is glad they caught it early. Same idea here: find the runaway while it's small and cheap. The prep is one command. Read more →
How is it priced?
Free is one developer, one per-session cap that blocks, email alerts and 7 days of replay, for as long as you like. Solo is $29 a month (or $290 a year): every budget at any scope, Slack, production stop-limits, 90-day replay. Team is custom pricing that covers everyone. Every workspace starts with a 14-day Solo trial, no card, then drops to Free. Never priced on your spend. Read more →
Get started
Stop reading invoices. Start reading flight logs.
Free for one developer, no card. One command, then sign in.
npm i -g tokenoscopy && tokenoscopy initRolling it out to a team? Leave your email and we'll set up a call.
For agents
/llms.txt/docs/changelogOr flip the HUMAN / MACHINE switch at the bottom of the screen.