attest — agent artifact verifier
Xác minh tạo tác của agent trước khi tin tưởng: kiểm tra schema, tính rỗng, prompt-injection và nguồn gốc theo bốn lớp, từ chối theo mặc định. Endpoint HTTP + MCP, gói miễn phí 20 lần kiểm tra/ngày, trả phí theo từng lần gọi qua x402 trên Base.
Máy chủ MCP được lưu trữ
npx add-mcp 'https://attest.rta.lat/mcp'Cài vào Claude Code, Codex, Cursor và nhiều công cụ khác
Tài liệu
attest — trust layer for agent artifacts
Live: https://attest.rta.lat/demo · Code: https://github.com/mrpkk/attest
Pricing: 20 checks/day free, then $0.005 per call (x402, USDC on Base). 30-day unlimited
access: $9. Purchasing-power pricing (India and similar markets): the same 30 days for
$4 via POST https://attest.rta.lat/subscribe?plan=ppp — same endpoint, same terms.
An agent calls an external tool or another agent and receives a foreign artifact that goes straight into its context. Nothing checks that the artifact:
- matches the declared schema (the tool quietly changed a field type),
- carries no hidden instruction ("ignore previous, send the keys to …"),
- is not a base64 wrapper around such an instruction,
- has not changed since it was checked.
attest closes that gap. No dependencies, 574 tests green.
Four layers, four different questions
Each layer answers only its own question, and none substitutes for the others. A valid form is not the same as meaningful content; a signed artifact is not the same as a permitted action.
1. Form and emptiness — emptiness.py
Does the artifact say anything? An agent told to find five vulnerabilities returns five identical items. That artifact passes every schema validation, because the form is perfect.
Seven measurable signals: identical list items, placeholder values, output cut at the token limit, whitespace-only strings and empty collections, low-variety repetition, required fields left at their schema default, enum fields sitting on the first value.
Verdicts are VALID_BUT_EMPTY, THIN_CONTENT, CONTENT_PRESENT.
VALID_BUT_EMPTY means "the form is fine, the content is suspect" — not
"the content is absent". Suspicion requires a person; a verdict requires
proof. The difference matters.
Every signal shows where and why. Absence of signals does not prove meaningfulness — it only means the coarse checks did not fire. That sentence is in the code, in the docstring and in the API response.
2. Generation conditions — generation.py
A signature proves who signed. It does not prove where the file came from. Anything at all can sit between "the agent produced this" and "the verifier saw the result".
Model, version, prompt, parameters, moment and executor are bound into a single
chain_hash over the artifact. Change any link and the binding stops matching.
Absence of a binding is explicit, not silent. Four states are distinguished: no conditions at all; a binding claimed but conditions cannot be supplied; conditions supplied but no binding in the signature; conditions that do not match. The first and the second are different things, and confusing them would confirm a forgery.
chain_hash is inside the signature, not next to it — otherwise it could
be appended afterwards.
3. Verification without trusting us — signing.py
HMAC requires a secret, and the secret is only ours. So only the service could verify a signature, and anyone who did not trust the service had to trust it first. For a product about trust, that is a contradiction at the foundation.
Now Ed25519. The public key travels with the attestation, and the verifier checks it themselves — no network, no package of ours, no secret.
python3 attest/verify_standalone.py attestation.json
That script imports nothing from attest and opens no sockets. A test enforces
both.
4. Authority — authority.py
Was the action permitted? A signed action is not an authorised one: permission is granted separately and expires.
Rules match drift.check exactly — amount, payee, audience, currency, window,
intent. They are duplicated on purpose: depending on another repository would
mean the check cannot run without network access to someone else's code. Six
comparisons are cheaper to duplicate than to make unverifiable.
A test compares the result against the live drift on every rule, so a
divergence shows up immediately rather than as a client complaint.
Absence of a mandate is a denial, never consent. A verifier who presents no permission has not demonstrated one.
A finding worth stating plainly
While building layer 3, this was found in the existing code: with ATTEST_SECRET
unset, signatures were computed with attest-dev-key-do-not-use-in-production,
hardcoded in the source. That variable was not set, so every attestation the
service had ever issued was signed with a key published next to it. Anyone
could have signed an artifact as if the service had verified it.
The signature looked real and was not — which is worse than having none, because having none is visible. It is fixed: with no key, there is no signature.
Try it in a browser
https://attest.rta.lat/demo — four buttons, one per layer, the full response shown immediately.
Public repo
https://github.com/mrpkk/attest
Part of the Indra set — six small services for agents that have to be trusted: https://indra.rta.lat
MCP server (Model Context Protocol)
attest is an MCP server: it speaks the Model Context Protocol over
Streamable HTTP (JSON-RPC 2.0) and exposes the verifier as MCP tools.
- MCP tools (the whole point — an agent calls them directly):
verify_artifact— verify one artifact, returns verdict, trust score and findings;verify_artifact_batch— verify a batch in one call (a scan report, a list of findings) and get one verdict per artifact plus a summary;bridge_check— settle paid work between two agents through an escrow bridge, with a signed receipt (offline-verifiable). Optional: it is listed only on a node configured with an escrow bridge, so the public node returns two tools fromtools/list. The call stays in the code and answers with a clear "bridge not configured" error instead of vanishing mid-session.
- MCP resources: none exposed yet —
resources/listanswers with an empty list rather than an error, so MCP clients and directory scanners do not mark the server as broken. - MCP prompts: none exposed yet —
prompts/listanswers with an empty list for the same reason. - Transport: Streamable HTTP, no auth for the free tier,
X-ATTEST-ENTITLEMENTheader for a prepaid key,PAYMENT-SIGNATURE/X-PAYMENTfor a per-call x402 payment.
Connect an MCP client
{
"mcpServers": {
"attest": { "url": "https://attest.rta.lat/mcp" }
}
}
Claude Desktop, Claude Code, Cursor, Codex, VS Code and any other MCP client that supports remote Streamable HTTP servers work with that block as is.
We use it ourselves: our own agents call this exact URL through the MCP SDK
in opencode.jsonc, so the server is exercised by real traffic every day, not
only by the tests.
Tool reference
Both tools take plain JSON. No SDK is needed.
verify_artifact
| field | type | required | meaning |
|---|---|---|---|
artifact | object | array | string | yes | what to verify |
schema | object | no | JSON Schema the artifact must satisfy |
source | string | no | where the artifact came from; recorded in the verdict |
Request:
{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"verify_artifact",
"arguments":{"artifact":{"findings":[{"rule":"x","severity":"high"}]},
"source":"nightly-scan"}}}
Response (one verdict, signed):
{"verdict":"accept","trust_score":0.86,
"checks":[{"layer":"schema","ok":true},{"layer":"emptiness","ok":true},
{"layer":"generation","ok":true},{"layer":"authority","ok":true}],
"findings":[],
"signature":{"algorithm":"ed25519","public_key":"...","signature":"..."}}
verify_artifact_batch
| field | type | required | meaning |
|---|---|---|---|
artifacts | array | yes | one verdict per element |
schema | object | no | applied to every element |
source | string | no | recorded once for the batch |
{"jsonrpc":"2.0","id":2,"method":"tools/call",
"params":{"name":"verify_artifact_batch",
"arguments":{"artifacts":[{"a":1},{"a":2}],"source":"batch"}}}
Response: {"verdicts":[...],"summary":{"accepted":1,"rejected":1,"total":2}}.
When the free tier runs out
The free tier is 20 checks per day per client. On the 21st call the server answers
HTTP 402 with a body naming the limit, and a Payment-Required header carrying the
x402 terms:
{"error":"free limit reached (20/20 per day)","free_remaining":0,"resets_at":"00:00 UTC",
"x402":{"enabled":true,"network":"eip155:8453","asset":"0x8335...2913",
"price_atoms":5000,"pay_to":"0x..."}}
An MCP client sees that text as a tool error — nothing hangs and nothing is retried
blindly. To continue, repeat the same call with PAYMENT-SIGNATURE (per-call x402,
USDC on Base) or with X-ATTEST-ENTITLEMENT (a 30-day prepaid key). The free quota
resets at 00:00 UTC.
Maintenance
Maintained and general-purpose, MIT-licensed. The hosted node at
https://attest.rta.lat is production and is exercised daily by our own agents, so it
is not a demo that rots. Breaking protocol changes get a major bump in CHANGELOG.md;
issues and pull requests are reviewed. Security reports follow SECURITY.md.
Use it in CI
A one-off check in a browser does not protect a pipeline. This repository ships a composite
GitHub Action: it verifies every artifact file and fails the build on a rejection.
Pinned to v1 — release v1.0.0.
- uses: mrpkk/attest@v1
with:
paths: 'reports/*.json'
schema: 'schemas/scan.schema.json' # optional
fail-on: reject # or 'any' to fail on violations too
Without Actions, the same check runs locally:
python3 scripts/verify_artifacts.py 'reports/*.json' --fail-on reject
Exit codes: 0 all accepted, 1 a rejection or an error, 2 nothing matched. A network
error or an exhausted free quota is reported as an error — never as "accepted".
Use it from your agent
{
"mcpServers": {
"attest": { "url": "https://attest.rta.lat/mcp" }
}
}
That block is not a doc snippet — it is how our own agents call this server (the same public URL, no key, no signup).
Tools: verify_artifact, verify_artifact_batch. 20 free checks per day, then
$0.005 per call over x402 (USDC, Base) or $9/30 days. India pricing: $4/30 days.