SiGR™ makes every AI inference call cryptographically accountable. It is a post-quantum signed receipt that proves exactly what the model saw, when, and which backend signed it. Anyone can check it. No one can fake it.
Four properties bound inside one signature. Change any field and the proof breaks. No NDA, no shared secret, no trust required.
Each product below emits a post-quantum signed receipt you can verify offline with no shared secret. Four are minting live receipts today at signer.thehiveryiq.com. The other five are built, tested, and benchmarked on hive-verifier-api and waiting on a public mint key before they go live. Every curl below matches the real, implemented schema. Verification is always free.
A signed, ordered, tamper-evident trail of an entire agent run. Every step, system, reasoning, tool call, final answer, gets hashed into a step-sealed chain and bound under one signature. Change any step, reorder them, or drop one, and the chain root stops matching. It is the record of what the agent saw and did before it acted, not something rebuilt after.
# sign a multi-step agent run. returns an envelope you can verify curl -s -X POST https://signer.thehiveryiq.com/sigr/chain \ -H 'Content-Type: application/json' \ -d '{"run_id":"run-1","tenant_id":"acme","agent_ref":"research-agent-v2", "final_step_index":2,"steps":[ {"step_index":0,"kind":"system","content_sha256":"13f92ad848457a123fec840d19215d17e9ad2be8ac646916377da068b5309712"}, {"step_index":1,"kind":"tool_call","content_sha256":"140a4e394d8a8df15251a360c6ee0ded2f0bbf10b5ae3a40f7875df6a01dd6f6","tool_ref":"search-tool"}, {"step_index":2,"kind":"final","content_sha256":"e5a3db2139b7e5d95d3e4b9454746d04fa50291a9a425490727b3b19054b4840"}]}' # verify it for free, no secret needed. paste the envelope returned above curl -s -X POST https://signer.thehiveryiq.com/sigr/chain/verify \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Binds the model, the exact input and output token counts, and the price schedule into one verifiable receipt per request. The bill stops being "trust our meter" and becomes a fact the customer, an auditor, or a regulator can check on their own. Silent model or quantization swaps get caught.
# sign an inference cost / billing receipt curl -s -X POST https://signer.thehiveryiq.com/sigr/bill \ -H 'Content-Type: application/json' \ -d '{"request_id":"r1","tenant_id":"acme","model_id":"llama-3.1-70b", "input_tokens":100,"output_tokens":50, "price_input_micro_usd_per_1k":50,"price_output_micro_usd_per_1k":150}' # verify it for free curl -s -X POST https://signer.thehiveryiq.com/sigr/bill/verify \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
A signed SLA bond. The latency ceiling, uptime floor, and penalty rate get locked into terms up front, and every measurement against them is signed. When the service misses, the breach and the penalty are recomputed from the recorded terms and measurement, not renegotiated after the fact. The penalty is capped at 1e12 micro-USD by schema, so a bad measurement can never mint an unbounded claim.
# sign an SLA bond: terms plus one measurement against them curl -s -X POST https://signer.thehiveryiq.com/sigr/bond \ -H 'Content-Type: application/json' \ -d '{"bond_id":"bond-1","tenant_id":"acme","customer_ref":"acme-prod", "terms":{"latency_ceiling_ms":500,"uptime_floor_ppm":999000, "penalty_rate_micro_usd_per_breach":5000000}, "measurement":{"observed_latency_ms":320,"observed_uptime_ppm":999600}}' # verify it for free curl -s -X POST https://signer.thehiveryiq.com/sigr/bond/verify \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Signs the route of a multi-model panel: which models voted, what each produced, the method, and the winner, all into one receipt. When a panel of models decides something that matters, the decision becomes a signed, replayable fact instead of a black-box vote no one can check.
# sign a model-panel route curl -s -X POST https://signer.thehiveryiq.com/sigr/consensus \ -H 'Content-Type: application/json' \ -d '{"panel_id":"p1","tenant_id":"acme","method":"majority","members":[ {"model_id":"gpt-4o","output_digest":"aa","seq":0}, {"model_id":"claude","output_digest":"aa","seq":1}, {"model_id":"llama","output_digest":"bb","seq":2}]}' # verify it for free curl -s -X POST https://signer.thehiveryiq.com/sigr/consensus/verify \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Public key for offline verification of the four live products above: GET /pubkey on signer.thehiveryiq.com.
These five ship the same signature scheme, the same open verify pattern, and the same honest boundary language as the four live products above, on thehiveryiq.com/v1 rather than signer.thehiveryiq.com. GET /schema/<dotted.type> for each of the five returns 200 in production today, and mint fails closed with 401 without a valid bearer token. A live schema and mint route prove the contract is deployed; they do not by themselves prove an authenticated production mint has been run for every caller. They are implemented, unit tested, adversarially tested, smoke tested, and benchmarked in hive-verifier-api.
Grounded Claim Attestation. Binds an answer's individual claims to their supporting evidence, each claim gets a hash, a support hash, and a support-strength score, folded into one claims_root. It tells you how many of an answer's claims were actually grounded, not whether the answer is correct.
# sign a grounded-claim attestation. GET /schema/sigr.gca is live in production curl -s -X POST https://thehiveryiq.com/v1/mint/sigr-gca \ -H 'Content-Type: application/json' -H 'Authorization: Bearer YOUR_MINT_TOKEN' \ -d '{"answer_id":"a1","tenant_id":"acme","method_hash":"27e74691455d8602859391429c52d4d584cd06cc95c076ae51c7fadb707ab61", "claims":[{"claim_id":"c1","claim_hash":"a6745486249786944a186ccc8110bf94aae5d86535116cc771dac71847504479", "support_sha256":"de5f86406b73a8d6cda558f0a577d3a302e8fa3692a9792d89fd296bd1ddde6c","support_strength_bp":8500}]}' # verify is open, no bearer token required curl -s -X POST https://thehiveryiq.com/v1/verify/sigr-gca \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Ground Truth Model anomaly signal. Watches for grounding anomalies, identity flicker, chain irregularity, cross-run divergence, and under-attested high-stakes output, then signs whether the combined signal crossed a declared trigger threshold. It reports a pattern was observed, never that anything is actually wrong.
# sign an anomaly-signal receipt. GET /schema/sigr.gitm is live in production curl -s -X POST https://thehiveryiq.com/v1/mint/sigr-gitm \ -H 'Content-Type: application/json' -H 'Authorization: Bearer YOUR_MINT_TOKEN' \ -d '{"subject_id":"s1","tenant_id":"acme","trigger_bp":2000, "signals":{"grounding_anomaly":false,"identity_flicker":false, "chain_irregularity":false,"cross_run_divergence":false,"under_attested_high_stakes":false}}' # verify is open, no bearer token required curl -s -X POST https://thehiveryiq.com/v1/verify/sigr-gitm \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Seals a KV-cache prefix to a model and a token span at a point in time, and can chain to a named parent entry. It attests the prefix and span were sealed as recorded. It does not attest the cached content is correct or that any later inference actually reused it.
# seal a cache prefix. GET /schema/sigr.cachesign is live in production curl -s -X POST https://thehiveryiq.com/v1/mint/sigr-cachesign \ -H 'Content-Type: application/json' -H 'Authorization: Bearer YOUR_MINT_TOKEN' \ -d '{"cache_id":"c1","tenant_id":"acme","model_id":"llama-3.1-70b", "prefix_sha256":"8a850cb21ac00f9b96a5189c314af16bc7917c88e03aff911ce07959e34f88e9", "token_span":{"start":0,"end":512}}' # verify is open, no bearer token required curl -s -X POST https://thehiveryiq.com/v1/verify/sigr-cachesign \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Binds a model identity to its weights hash, config hash, and serving endpoint, recomputing one manifest_sha256 over all four. It attests the manifest hash recomputes exactly from what is recorded. It does not attest the endpoint is currently serving those weights or that the config is safe.
# sign a model manifest. GET /schema/sigr.manifest is live in production curl -s -X POST https://thehiveryiq.com/v1/mint/sigr-manifest \ -H 'Content-Type: application/json' -H 'Authorization: Bearer YOUR_MINT_TOKEN' \ -d '{"manifest_id":"m1","tenant_id":"acme","model_id":"llama-3.1-70b", "weights_sha3":"9a6ed6149abf11ef6feecd89321edf74d8fc2c5c6c03c0c4c1146da75d44c206", "config_hash":"09a38eca0c750dd77c207b83cc398cd878c9d776c4ffac1d72f706f2d55e718a", "endpoint":"https://inference.acme.com/v1/generate"}' # verify is open, no bearer token required curl -s -X POST https://thehiveryiq.com/v1/verify/sigr-manifest \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Model Identity Route. Chains a sequence of model-identity steps, each with its own weights hash, config hash, and endpoint, folded into one lineage_root, and checks the final step matches an expected model. It attests the recorded lineage folds to that root. It does not attest any step actually ran.
# sign a model-identity lineage. GET /schema/sigr.mir is live in production curl -s -X POST https://thehiveryiq.com/v1/mint/sigr-mir \ -H 'Content-Type: application/json' -H 'Authorization: Bearer YOUR_MINT_TOKEN' \ -d '{"subject_id":"s1","tenant_id":"acme","expected_model":"llama-3.1-70b", "steps":[{"seq":0,"model_id":"llama-3.1-70b", "weights_sha3":"3eba85b19bc9708ca2e3e62b0d59c9ceb041860882bbc10a14391e8db61f98e2", "config_hash":"0ffa6f6ccd511b1b501fb38db1ef2a56f5d940a37a4b1cfce7019e26406fe0a2", "endpoint":"https://e0.acme.com"}]}' # verify is open, no bearer token required curl -s -X POST https://thehiveryiq.com/v1/verify/sigr-mir \ -H 'Content-Type: application/json' -d '{"envelope": ENVELOPE_FROM_ABOVE }'
Schemas for all nine types: /.well-known/schemas/ · example receipts: canon/examples on GitHub.
SiGR-Chain, SiGR-Bill, SiGR-Bond, and SiGR-Consensus above are live on the signing endpoint shown in the curl examples. The remaining five constituents, SiGR-GCA, SiGR-GiTM, SiGR-CacheSign, SiGR-Manifest, and SiGR-MiR, are also live now, deployed on the unified Hive typed receipt service at thehiveryiq.com/v1, unit tested, smoke tested, and benchmarked. Every schema below is public today, and every mint route above answers in production, gated by a bearer token.
Schema (all nine): sigr-chain-v1.json · sigr-bill-v1.json · sigr-bond-v1.json · sigr-consensus-v1.json · sigr-gca-v1.json · sigr-gitm-v1.json · sigr-cachesign-v1.json · sigr-manifest-v1.json · sigr-mir-v1.json
Source and mint/verify route docs: hive-verifier-api/README.md
Example receipts, signed with a published, non-production key so anyone can check the format: canon/examples/
Open, no-authentication verifier for any Hive receipt: thehiveryiq.com/verify/
Authenticated minting on the open Hive typed receipt service requires a bearer token Hive controls; verification is always open and free, with no token and no shared secret.
When an enterprise customer, an auditor, or a regulator asks "what did the model actually see, and is this the model we paid for?", logs are not proof. Logs can be edited. A signature can't.
The first platform that emits SiGRs makes the ones that don't look negligent by comparison.
Checking a receipt is free forever, for anyone, offline, no secret needed. You pay only to sign, and only past the free tier. One receipt equals one Bill receipt, one Bond measurement, one Chain step-sealed run, or one Consensus panel. Roll-ups are free. The metered fee gets you in the door, the license is the real deal. Usage is uncapped, with no monthly minimum and no ceiling. You pay for exactly what you sign and verify.
The table below shows the default Standard signature at each volume band. Need post-quantum (PQ) or anchored (MAX) receipts? The same volume discount applies on a higher base. See the full signature × volume matrix.
| Tier | Monthly signed receipts | Per receipt (Standard) |
|---|---|---|
| Build | first 1M | Free |
| Scale | 1M to 50M | $0.00027 |
| Platform | 50M to 1B | $0.00024 |
| Hyperscale | 1B+ | $0.00021 floor |
The price is graduated: each band is priced at its own rate, like cloud egress. Rates shown are the Standard signature tier. PQ and MAX carry the same volume discount on their higher base. The Hyperscale floor is negotiable. Trust-tier add-ons and per-product, suite, or sovereign licenses are below.
Drag to your monthly inference volume. This is the revenue line SiGR adds to your platform, and the cost line you bill through to enterprise customers who demand proof.
The base receipt is post-quantum and verifiable by anyone. For buyers who need entropy you can prove, time you can anchor, or signing rooted in silicon, three add-ons stack onto any SiGR product.
Quantum-derived randomness bound into the signing material, so the entropy behind every receipt is itself provable. Available per-receipt as a premium, or as a sovereign / on-prem custody license for buyers who must hold their own key material.
Anchors the receipt's time in a checkable chain of custody: capture, soft-commitment, then L1 finality on Base. Choose soft-confirm for speed, or full L1 anchoring when the timestamp itself has to be unforgeable.
Signing rooted in attested hardware with no software fallback. Available today as a software-attested backend under license; dedicated silicon acceleration is in development and not yet generally available.
Per-receipt pricing gets you in. When SiGR becomes load-bearing for your platform, embedded at the edge, in every enterprise RFP answer, the annual license is how serious platforms own the rail. Metered usage credits against the license.
Three paths, one rail. Start signing in under a minute on the free tier, meter on USDC over Base when you cross 1M receipts, or book a license when SiGR becomes load-bearing.
First 1M signed receipts every month, free. Instant key, usable immediately against every live endpoint above. No card, no call.
Past the free tier, metered at the graduated ladder and settled in USDC on Base, the same rail Hive uses for nano-payments. No invoices, no minimums. You pay only for what you sign.
See the ladderSingle-product, suite, or sovereign on-prem. Tell us your volume and jurisdiction and we will scope the deal. Metered usage credits against it.
SiGR is not a feature you bolt on. It is the trust and accountability layer the platforms serving inference at scale will be forced to adopt to stay credible. Here's the revenue at stake as it spreads.
A drop-in on your existing OpenAI-compatible API. Customers opt in per call. Zero change to your inference path beyond the ~7ms signing step, which fails closed.
A regulated buyer chooses you over a hyperscaler specifically because you can prove what you served and what you charged. Trust becomes your differentiator, not your liability.
That win puts "can you emit verifiable receipts?" into every subsequent RFP. Your competitors now face a question they can't answer, and the category tips.
You're not the platform scrambling to add accountability after an incident. You're the one who defined what accountable inference looks like. Inevitable, and yours.
A 30-day pilot on one model family. We instrument your API edge, you emit live SiGRs on opt-in traffic, and we put a verifiable receipt in front of your most demanding enterprise prospect. If the bill, the provenance, and the timeline don't verify in front of them, you owe us nothing.