# QED Proof > QED Proof is an independent verifier for AI agents' work. When an agent claims it did something, QED Proof reads the destination system itself — never the agent's own report — decides a verdict (Verified, Failed, Mismatch, Late, or Couldn't check), issues an Ed25519-signed receipt, appends it to an append-only Merkle log, and anchors the log on Base Sepolia (testnet) through EAS. Anyone can check a receipt with the open-source checker and a public RPC, without trusting QED Proof. Positioning: credit scores for AI agents, built on proof instead of reviews. ## Key facts - Evidence comes from the destination, never from the agent's report. - A verifier error, timeout, missing permission or unsupported action gives Couldn't check, never Verified. - Receipts are append-only and store fingerprints and the facts read (a commit SHA, an HTTP status), not content. - Anchoring today is on Base Sepolia (Base's test network), every ten minutes, via the Ethereum Attestation Service; Base mainnet is the next step. - QED Proof's spec, verifiers, node and checker are open source (Apache-2.0) at https://github.com/Nuraveda-Labs/qed-proof-core; the spec's technical name is Proof of Agent Work (poaw/0.1). It is self-hostable with docker compose. - QED Proof has no token and nothing to mine or buy. It is unrelated to token-mining "proof of work" projects with similar names; "Proof of Agent Work" (PoAW) is only the technical name of its open receipt spec. - QED Proof is not a code reviewer: it checks outcomes, not code quality. - Claims are sent today with POST /v1/claims and an API key. A Python SDK, an MCP tool and a Claude Code hook are planned, not built. - Pipelines (user-defined proof, e.g. prove only ad-budget changes and alert on unreported ones) are designed, not live. - Public signing keys: https://api.qedproof.site/.well-known/poaw-keys.json - Pricing (https://qedproof.site/pricing/): Free includes 1,000 receipts a month; Team $49 a month with 25,000 receipts, then $1 per 1,000; Scale $299 a month with 250,000 receipts, then $0.60 per 1,000. Yearly billing is two months free, every plan gets the same signed receipts, log and anchoring, and Enterprise is priced on request. ## Integrations - **GitHub** (verified today) — actions: `github.commit.push`, `github.pr.open`, `github.checks.pass`. The repository itself, through the QED Proof GitHub App with read-only permissions: is the commit on the branch, is the pull request open, did the check runs on that commit pass. - **Any URL** (verified today) — actions: `http.url.status`. The URL, fetched from our network at a pinned address, for the status the agent claimed. No connection needed. - **Meta Ads** (connect read-only today, verifier next). The ad account's own activity log — budget and bid edits with old and new values, and which app made them. Connect read-only (ads_read) today. - **Facebook Pages** (connect read-only today, verifier next). The Page's own published posts. Connect read-only today; the post verifier is next. - **Instagram** (connect read-only today, verifier next). The account's own published media, including Reels. Connect read-only today; the verifier is next. - **Slack** (verified today) — actions: `slack.message.post`. Messages the agent says it posted, read from the channel itself with a read-only user token: is there a message at that timestamp, and does its text match when the agent says what it wrote. - **X** (verified today) — actions: `x.post.publish`. The post by its ID: does it exist, did it come from the connected account, and does its text match when the agent says what it wrote. - **WhatsApp** (not started). Messages sent from a business number. Waiting on Meta's Embedded Signup. - **Gmail** (not started). Whether the email really left the mailbox. Gated on a connector catalogue and pricing check. - **GitLab** (connect read-only today, verifier next). Connect GitLab accounts read-only today. Verifiers for commits, merge requests and pipelines — the same checks as GitHub — are next. - **Google Ads** (not started). The account's change history (change_event), after Meta Ads — the second connector pipelines are designed around. ## Pages - [QED Proof](https://qedproof.site/): What QED Proof is and how a claim becomes a receipt. - [How it works](https://qedproof.site/how-it-works/): The five steps from an agent's claim to an anchored receipt, with the claim payload. - [Integrations](https://qedproof.site/integrations/): Which destinations are verified today, which connect with verifiers next, and which aren't started. - [Use cases](https://qedproof.site/use-cases/): Coding, ops and marketing agents — and what is verified for each today. - [Pricing](https://qedproof.site/pricing/): Free, Team and Scale plans, priced per receipt, with proof never paywalled. - [Contact](https://qedproof.site/contact/): Talk to the team about sales, enterprise, partnerships, security or press. - [Open source](https://qedproof.site/open-source/): Proof of Agent Work: the Apache-2.0 spec, verifiers, node and checker, and how to self-host. - [FAQ](https://qedproof.site/faq/): Plain answers on verdicts, trust, data, integrations and anchoring. - [Security](https://qedproof.site/security/): What QED Proof stores, what it can touch, and how to check it. - [Blog](https://qedproof.site/blog/): Notes on verifying what AI agents actually did. - [Glossary](https://qedproof.site/glossary/): Plain definitions of the terms behind verifying AI agents, and how QED Proof implements each. ## Use cases - [Coding agents](https://qedproof.site/use-cases/coding-agents/): Verified today: commits, pull requests and CI results on GitHub. - [Ops agents](https://qedproof.site/use-cases/ops-agents/): Verified today: URLs answering as claimed, and the CI run behind the deploy. - [Marketing agents](https://qedproof.site/use-cases/marketing-agents/): Partly verified. Posts on X and messages in Slack are verified today. Meta (Pages, Instagram, Ads) connects read-only now; its verifiers for posts, Reels and ad-budget changes are next, and until then a claim against Meta returns Couldn't check. ## Blog - [Your agent says it's done. That isn't evidence.](https://qedproof.site/blog/agents-say-done-is-not-evidence/): An AI agent's success report is the one source that can't settle whether the work landed. Why QED Proof reads the destination instead, and why it fails toward "Couldn't check". - [How to check a QED Proof receipt without trusting us](https://qedproof.site/blog/check-a-receipt-without-trusting-us/): A QED Proof receipt carries three independent proofs — a signature, a log inclusion proof and an on-chain anchor. Here is what each one proves, and how to check all three yourself. - [Pipelines: prove only what costs you money](https://qedproof.site/blog/pipelines-prove-what-costs-you-money/): Proving that an agent pushed a commit keeps nobody up at night. Proving that it changed an ad budget does. Pipelines — designed, not yet live — let you define which actions deserve a receipt. ## Glossary - [Agent claim](https://qedproof.site/glossary/agent-claim/): A structured statement from an AI agent that it performed an action on a named target — the untrusted input QED Proof verifies. - [Agent reputation](https://qedproof.site/glossary/agent-reputation/): A track record for an AI agent computed only from verified outcomes rather than reviews or self-report — the "credit score for agents". QED Proof's score isn't built yet. - [AI agent verification](https://qedproof.site/glossary/ai-agent-verification/): Independently confirming that an AI agent's claimed work actually happened, using evidence the agent didn't produce — and making the result checkable by others. - [Append-only log](https://qedproof.site/glossary/append-only-log/): A log that can only grow — entries are added, never edited or removed — and that can prove it. Every QED Proof receipt is written to one. - [Content fingerprint](https://qedproof.site/glossary/content-fingerprint/): A hash of normalised content that proves what was said without storing what was said — how a QED Proof receipt checks content while keeping none of it. - [Couldn't check (unverifiable)](https://qedproof.site/glossary/unverifiable/): The verdict given when a verifier can't determine what happened — never evidence for or against the agent, and never silently upgraded to Verified. - [Destination verification](https://qedproof.site/glossary/destination-verification/): Checking an AI agent's work by reading the system the work was meant to land in, rather than the agent's own logs or report. - [Ed25519](https://qedproof.site/glossary/ed25519/): A fast, widely used digital-signature scheme (EdDSA over Curve25519, RFC 8032) with 32-byte public keys and 64-byte signatures. QED Proof signs every receipt with it. - [ERC-8004](https://qedproof.site/glossary/erc-8004/): A draft Ethereum standard for trustless AI agents, with registries for agent identity, reputation and validation. QED Proof plans to post its verdicts to the validation registry; that isn't built yet. - [Ethereum Attestation Service (EAS)](https://qedproof.site/glossary/ethereum-attestation-service/): An open protocol for recording signed statements, or attestations, on Ethereum and its layer-2 networks against registered schemas. QED Proof uses it to anchor its log on Base Sepolia (testnet). - [False completion claim](https://qedproof.site/glossary/false-completion-claim/): When an AI agent reports that a task is done but the outcome never happened at the destination — the failure QED Proof exists to catch. - [Inclusion proof](https://qedproof.site/glossary/inclusion-proof/): The short list of sibling hashes that proves one entry is in a Merkle tree with a given root — how a receipt proves it's really in the log. - [JSON Canonicalization Scheme (JCS)](https://qedproof.site/glossary/json-canonicalization-scheme/): RFC 8785 — a rule for serialising any JSON value to exactly one byte sequence, so that it can be hashed and signed reproducibly. - [Merkle tree](https://qedproof.site/glossary/merkle-tree/): A tree of hashes in which each parent hashes its children, so one root hash commits to every item beneath it — the structure behind QED Proof's receipt log. - [On-chain anchoring](https://qedproof.site/glossary/on-chain-anchoring/): Publishing a log's root hash to a public blockchain so it's fixed in time and can't be backdated — QED Proof anchors its receipt log on Base Sepolia (testnet) through EAS. - [Proof of Agent Work (PoAW)](https://qedproof.site/glossary/proof-of-agent-work/): The technical name of QED Proof's open receipt spec — the format, verifiers, node and checker, published under Apache-2.0. Not a token, and unrelated to proof-of-work mining. - [Reason code](https://qedproof.site/glossary/reason-code/): A machine-readable code on a verdict that says exactly why it came out that way — for example not_found, target_mismatch or permission_denied. - [Trust levels](https://qedproof.site/glossary/trust-levels/): The four levels of assurance a Proof of Agent Work receipt can reach — Signed, Anchored, Attested and Web-proven. QED Proof receipts reach level 2 today. - [Unclaimed change](https://qedproof.site/glossary/unclaimed-change/): A change at a destination that no agent claim explains — an agent that acted and didn't say so. Detecting them is part of QED Proof's pipelines design, which isn't live yet. - [Verdict](https://qedproof.site/glossary/verdict/): The outcome of verifying an agent claim — Verified, Late, Mismatch, Failed, or Couldn't check — decided from what the destination shows. - [Verification receipt](https://qedproof.site/glossary/signed-receipt/): A signed, tamper-evident record of one verification — the claim, what the destination showed, and the verdict — that anyone can check without trusting the issuer. - [Verifier profile](https://qedproof.site/glossary/verifier-profile/): The versioned, published rules for checking one kind of action — what to read at the destination, how to compare it, and how long to wait. ## Optional - [Full text for language models](https://qedproof.site/llms-full.txt): the FAQ, the mechanism and every blog post in full. - [Proof of Agent Work on GitHub](https://github.com/Nuraveda-Labs/qed-proof-core): the spec, verifiers, node and checker.