
Tokscale Gateway
One endpoint for your AI traffic. Metered, paid, and recorded at the terminal boundary.
Proof of Actual Work
Pay per model request with x402, get a signed receipt for every call,
and turn verified usage into a Builder Passport on GIWA.
The proof loop
Four layers. Each one consumes the previous layer's output.
An OpenAI and Anthropic compatible gateway. Every request is paid on-chain via x402 and written to an immutable ledger, so usage stops being self-reported.
Every terminal request gets an EIP-712 signed receipt. Once per epoch, receipts fold into a single merkle root. 1M or 100M requests. Same on-chain cost.
Verified usage accrues into a portable work history: tokens, active days, and per-model skill signals, bound to a Dojang Verified Address as attestations.
GIWA ecosystem projects post bounties for verified builders. Payouts accrue back as more proof.
The stack
The gateway is the data engine. Identity is the moat. Earn is the payoff.

One endpoint for your AI traffic. Metered, paid, and recorded at the terminal boundary.

HTTP 402 ported to GIWA with a self-hosted facilitator and canonical Permit2 proxies. Credit top-up today, per-request settlement staged.

Verified usage accrues into a work history you can carry. Attestations about an address, never transferable tokens.
The missing trust layer
Every record keeps its source, parser, verification level, and privacy-preserving evidence.
01Local AI usage
Capture activity from supported AI coding tools without exposing private work.
02Canonical receipt
Bind tokens, cost, model, parser version, session root, and device context.
03Verification target
Preserve the evidence needed for future provider matching and challenge review.
04Network publication
Publish local evidence on Local Anvil today without presenting it as GIWA Sepolia trust.
Dojang · the seal
Identity lives as attestations about an address, never as tokens owned by one. Non-transferable by construction, with zero PII on-chain.
Verification tiers
Self-reported usage is forgeable. Verified usage is an asset. Tokscale separates the two so confidence, not just volume, shapes leaderboards and rewards.
Published activity without an independent evidence check.
Published identity and activity facts pass the checks named by their source.
Provider or billing records independently match the published activity.
A write-once gateway receipt records terminal observed usage.
A named network publication commits to the evidence root.

Open-source base
Tokscale already ships as an open-source CLI. That usage is self-reported and labeled that way; routing the same activity through the gateway is what makes it verifiable.
4.7K
GitHub stars
301.7K
npm downloads
42
client integrations
Reads Claude Code, Codex, and 40 more clients.
Built on GIWA
A hash alone does not make usage true. Publication is Local Anvil-only today; GIWA Sepolia hosting, source verification, challenge history, and identity authority remain separate prerequisites.

Tokscale Verified
Turn selected AI usage into verifiable developer reputation, without exposing prompts, source code, or private project content.