EVIDIQ Vault
Governed, tamper-evident memory and audit trail for autonomous AI agents, anchored on 0G. Record agent actions, hash-chain sequences, seal Merkle segments, redact sensitive payloads via tombstoning without breaking hash chains, and audit continuity across namespaces.
EVIDIQ Vault
Agent #9622Quickstart
Connect the Streamable HTTP MCP endpoint, inspect capabilities, and validate your record input before appending.
claude mcp add --transport http evidiq-vault https://mcp.evidiq.dev/vault/mcp
Or probe the live pricing discovery endpoint:
curl -s https://mcp.evidiq.dev/vault/x402
Prefer a Skill file? Fetch the agent-readable EVIDIQ Vault Skill:
curl -s https://mcp.evidiq.dev/vault/skill.md
What Vault is for
Append-Only Hash Chains
Monotonically sequenced record log where each record stores the SHA-256 hash of the previous record.
0G Merkle Sealing
Compute binary SHA-256 Merkle roots over record batches and anchor sealed segments to 0G Storage.
Retention & Tombstoning
Redact raw payload contents while preserving contentDigest and recordHash, ensuring chain integrity.
Secret Interception
Automatically reject private keys, mnemonics, and bearer tokens before they enter the record chain.
Continuity Audits
Evaluate namespace logs for INTACT, GAP, or BREAK verdicts with detailed policy violation tracking.
Inclusion Proofs
Generate and verify Merkle inclusion proofs for individual records within sealed segments.
Ten MCP tools
Five free tools support preflight, validation, and offline verification. Five paid tools handle durable append, queries, sealing, auditing, and retention.
Paid memory & audit tools
- append_record — Append an action/decision record to a governed namespace with SHA-256 hash chaining and secret detection. (0.005 USDT0)
- query_records — Query bounded records by actor, authority, action, tags, or window with optional Merkle inclusion proofs. (0.01 USDT0)
- seal_segment — Compute binary SHA-256 Merkle root across a range of records and optionally seal to 0G Storage. (0.015 USDT0)
- audit_report — Audit namespace continuity and output INTACT, GAP, or BREAK verdict with policy violation details. (0.02 USDT0)
- enforce_retention — Apply retention policy by tombstoning raw payload data while preserving contentDigest and recordHash linkage. (0.03 USDT0)
Free preflight and verification
- vault_capabilities — Inspect engine parameters, namespace pattern limits, secret scanning rules, and x402 pricing. (free)
- validate_record — Validate record input schema and test for secrets without appending or paying. (free)
- estimate_cost — Quote the immutable cost of any paid tool. (free)
- verify_chain — Verify cryptographic hash-chain continuity and genesis linkage across a set of records. (free)
- get_receipt — Retrieve a content-addressed JSON receipt or proof by exact ID. (free)
Engine limits
65536
16
500
^[a-z0-9][a-z0-9._/-]{2,63}$
Chain format v1: SHA-256 prevHash linkage, with an all-zero prevHash at genesis.
Recommended workflow
Settlement happens before a paid tool runs, so a malformed argument is still a paid call. Preflight for free first.
- Call vault_capabilities for current limits, secret-scanning rules, and prices.
- Call validate_record to check the record schema and secret scan without appending or paying. contentDigest must be bare 64-character SHA-256 hex — a 0x prefix is rejected.
- Call estimate_cost for the intended operation.
- Submit one paid call per request. Vault enforces strict x402 payment authorization.
- Re-read and verify afterwards for free with get_receipt and verify_chain.
Records and receipts
Every paid append returns the stored record, its content-addressed ID, and an integrity envelope carrying a SHA-256 digest and an EIP-191 signature.
{
"engine": "EVIDIQ-Vault/1.0",
"chainFormatVersion": "v1",
"data": {
"record": {
"seq": 1,
"namespace": "evidiq-sdk-proof",
"actor": "0xd6B658dC6e53444bF9Cba598aFdd21Ede0A62Fb9",
"authority": "operator",
"action": "verify",
"contentDigest": "18b7a227d4c939603fb44112e78be92e434e46cd6abf0ecd92e56e45746cd28d",
"prevHash": "0000000000000000000000000000000000000000000000000000000000000000",
"recordHash": "e81012620100231edcc2488d0571dc0586322e1fc0f95be0cc795b8b5e8aa5ac"
}
},
"integrity": { "algorithm": "SHA-256", "digest": "..." }
}verify_chain recomputes each recordHash and checks the prevHash linkage, so a reordered or edited log is detectable without trusting Vault. enforce_retention keeps contentDigest and recordHash when it redacts a payload, which is why continuity still verifies as INTACT after redaction.
x402 pricing
| Tool | Atomic | USDT0 | Access |
|---|---|---|---|
| append_record | 5000 | 0.005 | x402-paid |
| query_records | 10000 | 0.01 | x402-paid |
| seal_segment | 15000 | 0.015 | x402-paid |
| audit_report | 20000 | 0.02 | x402-paid |
| enforce_retention | 30000 | 0.03 | x402-paid |
| vault_capabilities | 0 | Free | Always ungated |
| validate_record | 0 | Free | Always ungated |
| estimate_cost | 0 | Free | Always ungated |
| verify_chain | 0 | Free | Always ungated |
| get_receipt | 0 | Free | Always ungated |
Payments use x402 v2 exact scheme with USDT0 (6 decimals) on X Layer (eip155:196). The public discovery endpoint lists all ten tools.
Verification and settlement run through the official OKX Onchain OS Payment SDK (@okxweb3/x402-core and @okxweb3/x402-evm). The OKX facilitator verifies each authorization and settles it on X Layer before any record is written. Each price reaches the SDK as an explicit USD₮0 atomic asset amount rather than a USD string, so neither the fee nor its token can be substituted by conversion.
Settled on X Layer
A live append_record call paid 0.005 USDT0 (5000 atomic) and stored seq 1 with an all-zero genesis prevHash. An audit_report call paid 0.02 USDT0 (20000 atomic) and returned verdict INTACT with no policy violations. Both receipts are status 0x1, and both were broadcast by an OKX facilitator relayer rather than a Vault-held key — the on-chain evidence that settlement ran through the official SDK. The appended record was then re-read with the free get_receipt, so the paid write is durable.
0x7e96398e1f2bae0637af884b55b35cbc82099b8241858a1fdde3ab67c94db4b60x2b67c2109fc799b9d167aa6ebd710b478a56c7ad9f026e07069febcd525bbb6eLicensing
EVIDIQ owns and licenses its original Vault code under MIT. Third-party dependencies maintain their own open-source licenses preserved in THIRD_PARTY_NOTICES.md.
Vault produces tamper-evident evidence of agent behaviour, not a judgement about it. Route dependency provenance to EVIDIQ Lineage and endpoint or Agent Skill security scanning to EVIDIQ Sentinel.