TokenBeaver sits between your agent and the model, trimming wasted tokens before they ever hit your bill. Local-first, so your prompts never leave your machine.
Install straight from the VS Code Marketplace, or download the .vsix and add it by hand — your call.
Three steps. One line of config. Your keys, your machine — TokenBeaver never sees your code or your account.
Point your editor at TokenBeaver's local endpoint. One line of config — no SDK, no rewrite.
We optimize both input and output, alongside context and cache. Every prompt stays on your machine.
A leaner request reaches the model with your own key. Same answer, fewer tokens, smaller bill.
Prompts, code, and keys never leave your machine. TokenBeaver optimizes in-process.
40–70% lower API spend, depending on session length, model, and workload.
Drops into Claude Code, VS Code, and others with a single endpoint change.
Bring your own provider keys. We never proxy billing or touch your model account.
15–25% more usable Claude Code quota in a 5-hour rolling window, depending on session length and number of sub-agents.
See every optimization applied to every request. Nothing hidden, nothing silently dropped.
For trying it out on a real project.
Add to VS CodeOr $100 / year — under $8.40 a month.
Start 14-day trialTwenty free optimizations per month, no card required. See the savings on your own project today.
One Pro plan if you want it. Free forever if you don't. No seats, no usage surprises.
For trying TokenBeaver on a real project, no card needed.
Add to VS CodeJust $8.33 / month, billed annually. Two months free vs monthly.
Billed monthly. Switch to annual any time to save 44%.
It sits between your editor and the model, stripping redundant context, deduping repeated content, and caching locally — so each request carries fewer tokens while returning the same answer.
No. TokenBeaver is local-first — optimization happens in-process on your machine, and requests go directly to the model with your own key.
One optimized request passing through the gate. Free gives you 20 per month; Pro is unlimited.
If you use Claude Code, Cursor, Cline, or any tool that lets you set a base URL, yes. Setup is a single endpoint change.
Yes — cancel any time in the 14 days and you won't be charged. Self-serve, no email required.
Pro unlocks unlimited optimizations right inside the extension you've installed. 14 days free — we'll only charge you when the trial ends, and you can cancel any time before then.
Install the extension, then point your editor at the local gate. No SDK, no code changes, no account hand-off — your keys stay yours.
Search TokenBeaver in the VS Code Marketplace, or install directly.
Click Start Gateway in the TokenBeaver panel, or open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) and run TokenBeaver: Start Gateway. Either way, the extension starts a local server and shows you a Base URL.
Paste the Base URL into your coding tool's API settings. Keep using your own API key — TokenBeaver forwards it unchanged.
We're developers who lean on AI coding tools every day. The output was worth it — the invoices, less so. Most of what we were paying for was redundant context shipped to the model over and over again.
So we built a small gate that sits between the editor and the model, trims the waste, and forwards a leaner request. It runs locally, uses your own keys, and never sees your code. In about a month of internal testing it's cut our API spend by 40–70% depending on the model — without changing a single answer.
TokenBeaver is independent and intentionally simple. No data harvesting, no lock-in, no surprise pricing. If it saves you money, keep it. If it doesn't, the free tier costs you nothing.
Your prompts, code, and keys stay on your machine. Privacy isn't a setting — it's the architecture.
One Pro plan, a real free tier, and no usage traps. You always know what you'll pay.
Every optimization is visible. We'd rather under-claim and show you the numbers.
No investors to please, no growth-at-all-costs. Just a tool that earns its keep on your machine.
Last updated: June 28, 2026
TokenBeaver is built local-first. The short version: your prompts, your code, and your API keys stay on your machine. We designed the product so that we never see them in the first place.
TokenBeaver is a VS Code extension and local gateway that sits between your AI coding agent and your model provider, trimming wasted tokens before requests are sent. Optimization happens in-process on your own device.
We never receive, store, or transmit any of the following:
TokenBeaver optimizes in-process on your own device. Nothing from your editor or model interactions is sent to TokenBeaver's servers — requests go directly from your machine to your chosen model provider using your own key, so we are never in the path of your billing relationship with that provider.
We don't ask you to create an account, and we don't collect your email or personal profile. To handle billing and basic product analytics, we collect only a limited set of data:
We use the data above to provide and secure the service, enforce free and Pro plan limits, process payments, respond to support requests, and understand aggregate product usage. We do not save or sell your data, and we do not use your code or prompts to train models.
We rely on a small number of processors to run the business — for example, Stripe for payments and standard infrastructure providers for hosting our website and account systems. Your model provider (such as Anthropic) receives your requests directly from your machine under your own agreement with them.
We keep billing records only as long as required for legal and accounting purposes. Since there's no account or email on file, the only data tied to you is your hashed device record. You can request deletion of your device record at any time by emailing [email protected]. Aggregate stats are anonymous and cannot be linked to a device.
We may update this policy as the product evolves; material changes will be reflected by the date above. Questions about privacy? Email [email protected].
Last updated: June 28, 2026
These terms govern your use of TokenBeaver — the VS Code extension, local gateway, and website. By installing or using TokenBeaver, you agree to them. If you're using it for an organization, you accept on its behalf.
TokenBeaver is a tool that optimizes AI coding requests locally on your device to reduce token usage. You bring your own model-provider API key; TokenBeaver forwards your requests to that provider unchanged except for the optimizations it applies. We don't resell model access and we aren't a party to your agreement with your model provider.
The Free plan includes 20 optimizations per month with full feature access. Pro is $15/month or $100/year and includes unlimited optimizations, billed through our payment processor. Pro starts with a 14-day free trial; if you don't cancel before it ends, the plan converts to paid at the price shown at signup.
Use TokenBeaver lawfully and in line with your model provider's terms. Don't attempt to resell, reverse-engineer for circumvention, abuse the optimization service, or use it to violate others' rights. We may suspend access for activity that threatens the service or other users.
You're responsible for your own API keys, your model-provider charges, and the code and prompts you run through your editor. TokenBeaver runs locally and uses a fail-open design, but you should keep your own backups and review outputs before relying on them. Don't use TokenBeaver to circumvent model-provider rate limits in ways that violate those providers' terms of service.
TokenBeaver is provided as-is. We're not liable for API costs, model behavior, or data loss. Token savings are not guaranteed — results vary by session length, model, workload, and number of sub-agents. To the extent permitted by law, our liability is limited to the amount you paid us in the 6 months before a claim.
We may update these terms as the product evolves; material changes are reflected by the date above and continued use means acceptance. Questions? Email [email protected].
Last updated: June 28, 2026
We want TokenBeaver to pay for itself. Between the free tier and the 14-day Pro trial, you can confirm it saves you money before you're ever charged — but if something's wrong, here's how refunds work.
The Free plan (20 optimizations per month) and the 14-day Pro trial let you evaluate TokenBeaver at no cost. No card is charged until a trial ends, and you can cancel any time before then to avoid being billed.
If you're charged for Pro and aren't happy, email us within 7 days of that charge and we'll refund it in full — no interrogation. This applies to your first paid term (monthly or annual).
For annual plans, the 7-day window applies from the date of each charge, including renewals. After the window, we don't provide prorated refunds for the remainder of a term, but you can cancel to stop the next renewal and keep access until the period ends.
Email [email protected] from the address on your subscription. Approved refunds are issued to your original payment method through Stripe and typically appear within 5–10 business days, depending on your bank.
Route Claude Code through TokenBeaver's local gateway to cut token usage — same answers, fewer tokens, and your API key never leaves your machine.
The TokenBeaver VS Code extension installed, and your own Anthropic API key set up for Claude Code.
Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) and run TokenBeaver: Start Gateway. The extension starts a local server and shows a Base URL such as http://127.0.0.1:4505/v1.
Set the base URL environment variable — note: no /v1 suffix here:
# add to your shell profile export ANTHROPIC_BASE_URL="http://127.0.0.1:4505"
Use Claude Code as normal. TokenBeaver shows the optimizations applied to each request. If the gateway is ever unavailable, requests fail open and go directly to Anthropic, so your workflow never breaks.
Seeing 404s? Make sure ANTHROPIC_BASE_URL has no trailing /v1. Port already in use? Change the gateway port in TokenBeaver settings and update the variable to match.
Payment confirmed and your license is being activated. Head back to VS Code — unlimited optimizations are now unlocked.
Confirm activation to use your TokenBeaver Pro license on this machine.
Having trouble? Email [email protected]
The beaver couldn't find what you're looking for — but your tokens are still safe.