EVIDIQ Redact
Deterministic detection and removal of sensitive data — before an agent sends anything anywhere. Redact finds sensitive values by structure rather than by guesswork, removes them in the mode you choose, and returns a signed report of exactly what was found and where. No model in the hot path, and nothing is stored.
EVIDIQ Redact
Agent #9700Quickstart
Connect the Streamable HTTP MCP endpoint, then preflight for free before paying.
claude mcp add --transport http evidiq-redact https://mcp.evidiq.dev/redact/mcp
Or probe the live pricing discovery endpoint:
curl -s https://mcp.evidiq.dev/redact/x402
Prefer a Skill file? Fetch the agent-readable EVIDIQ Redact Skill:
curl -s https://mcp.evidiq.dev/redact/skill.md
What Redact is for
Before a prompt
Strip customer records, credentials, and keys out of context before it reaches a model.
Checksum-validated detection
Luhn for cards, mod-97 for IBANs, wordlist plus checksum for BIP-39 mnemonics, EIP-55 for EVM addresses.
Credential interception
AWS, GitHub, OpenAI, Anthropic, Stripe keys and .env blocks by issuer shape.
Structure-preserving output
Markdown, HTML, CSV, JSON, NDJSON, and plain text keep their shape after redaction.
Policy verdicts
Versioned profiles return PASS, REVIEW, or BLOCK with the violations that caused it.
Zero retention
Content is processed in memory; artifacts live in a short-TTL cache addressed by digest.
Ten MCP tools
Five free tools cover discovery, preflight, and offline verification. Five paid tools handle detection, redaction, policy evaluation, and dataset de-identification.
Paid detection & redaction
- redact_text — Detect and redact one text blob; returns redacted text, findings with offsets, and a signed report. (0.005 USDT0)
- scan_document — Inventory only, no rewriting: which sensitive entities exist, where, and at what confidence. (0.01 USDT0)
- redact_document — Structure-preserving redaction of markdown, HTML, CSV, JSON, NDJSON, or plain text. (0.015 USDT0)
- policy_check — Evaluate content against a named policy profile and return PASS, REVIEW, or BLOCK with violations. (0.02 USDT0)
- deidentify_dataset — Column-aware pseudonymization with stable tokens and re-identification-risk warnings. (0.03 USDT0)
Free preflight and verification
- redact_capabilities — Detector catalog and versions, policy profiles, limits, redaction modes, and full pricing. (free)
- validate_input — Validate size, format, and encoding and return the finding count — without returning content or charging. (free)
- estimate_cost — Quote the immutable cost of any paid tool. (free)
- verify_redaction_report — Recompute the report digest and verify its EIP-191 signature. (free)
- get_artifact — Retrieve a redacted artifact by id, within its in-memory TTL. (free)
Engine limits
262144
600000 (10 minutes, in memory)
1.0.0 — 20 detectors
mask · label · hash · token · remove
Policy profiles: pii-basic, pci-lite, health-lite, crypto-hygiene, pre-prompt. Live values come from the free redact_capabilities tool.
Recommended workflow
Settlement happens before a paid tool runs, so a malformed argument is still a paid call. Preflight for free first.
- Call redact_capabilities for the detector set version, profiles, and prices.
- Call validate_input — it returns the finding count for free, so you can decide whether redaction is worth buying.
- Call estimate_cost for the intended operation.
- Submit one paid call per request. Redact enforces strict x402 payment authorization.
- Verify offline with verify_redaction_report, and re-fetch output with get_artifact while its TTL lasts.
What a report proves, and what it does not
A report proves that this exact input, under this detector set version and this policy version, produced this exact output and these findings. Anyone can re-run and compare digests. It does not prove the content is now free of sensitive data: deterministic detection has a recall limit, and unstructured personal names and postal addresses are the weak spot. They are reported at low confidence and can never alone drive a BLOCK.
Findings carry offsets and detector ids, never the matched values. A report that quoted what it found would itself be a leak, and reports get pasted into tickets.
{
"engine": "EVIDIQ-Redact/1.0",
"detectorSetVersion": "1.0.0",
"policy": "pre-prompt",
"findings": [
{ "detector": "BIP39_MNEMONIC", "family": "crypto_secret",
"confidence": "validated", "start": 812, "end": 913, "action": "remove" }
],
"verdict": "BLOCK",
"integrity": { "algorithm": "SHA-256", "digest": "...", "signature": "0x...", "signer": "0x8a3c7524..." }
}x402 pricing
| Tool | Atomic | USDT0 | Access |
|---|---|---|---|
| redact_text | 5000 | 0.005 | x402-paid |
| scan_document | 10000 | 0.01 | x402-paid |
| redact_document | 15000 | 0.015 | x402-paid |
| policy_check | 20000 | 0.02 | x402-paid |
| deidentify_dataset | 30000 | 0.03 | x402-paid |
| redact_capabilities | 0 | Free | Always ungated |
| validate_input | 0 | Free | Always ungated |
| estimate_cost | 0 | Free | Always ungated |
| verify_redaction_report | 0 | Free | Always ungated |
| get_artifact | 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 detection runs. 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 redact_text call paid 0.005 USDT0 (5000 atomic) on content carrying an email, a Luhn-valid card number, a BIP-39 mnemonic, and a private key. All four were detected, the verdict was BLOCK, every secret was removed from the output, and no matched value appeared in the report. The receipt is status 0x1 and the transaction was broadcast by an OKX facilitator relayer rather than a Redact-held key — the on-chain evidence that settlement ran through the official SDK. Test values were synthetic.
0x001261081770e0c9bc82a736a4a7e8739d25fe01ca578a445aa311f238e8efc4Licensing
EVIDIQ owns and licenses its original Redact code under MIT. Third-party dependencies maintain their own open-source licenses preserved in THIRD_PARTY_NOTICES.md.
Redact inspects what an agent emits. For what an agent consumes, route endpoint and Agent Skill scanning to EVIDIQ Sentinel; for durable action logging, use EVIDIQ Vault. No compliance certification is claimed; policy profiles are engineering defaults, not legal advice.