OpenWorkProof
You can use this MCP server to inspect, sign, verify, and export OpenWorkProof evidence, run the reference SOC workflow, and operate ledger-backed verification/acceptance workflows.
Fetch synthetic Slips alerts, analyze supplied JSONL into cases and a signed evidence bundle, and verify bundles with separately supplied trust pins; missing pins yield UNKNOWN.
Generate Ed25519 key pairs, compute key IDs and canonical digests, sign payloads, and verify signatures.
Verify WorkOrder identity bindings and nested AgentRequest/HumanDecision claims.
List canonical domains and validate profiles, scopes, observations, judgment commitments, and action-binding manifests; compare scopes with observations.
Replay SQLite ledgers; query status and binding status; explain binding decisions; derive settlement readiness.
Run ledger-coordinated tests and repo-read operations; commit verification results or signed decisions; prepare unsigned decision drafts.
Build public, diagnostic, or customer-private delivery packages; verify and render surface bundle reports.
Retrieve authoritative JSON Schemas and their frozen digests.
Analyze local or remote repositories for structure, languages, line counts, and dependencies.
Provides a GitHub Action that integrates OpenWorkProof verification and acceptance evidence into existing pull request delivery workflows.
OpenWorkProof
English · 简体中文
Contributors: dengyier & Tim Shing(成孝添)
Check what an AI agent was allowed to do, what it recorded, and whether its evidence holds up.
OpenWorkProof (OWP) is an open-source toolkit for developers and teams using AI agents. Use it to review signed work records, check exported evidence, or run a bounded Codex preview-editing task. Security engineers can also try a reference workflow for reviewing Slips alerts.
You keep the final decision to accept or reject the result. Connecting OWP to an agent gives it tools; protection of a Codex task requires a separate Guard setup.
Get started · Codex Guard · Docs · Community
Current release: 1.6.1 · PyPI · Release notes · MCP Registry
What can you do with it?
Your task | What OWP helps you do | Start here |
Review an agent's work records | Check signed task scope, recorded actions and evidence; export a bundle another reviewer can inspect. Your integration must supply the records. | |
Let Codex edit a preview copy | Explicitly allow supported text edits inside one preview directory and reject out-of-scope requests in the guarded launch path. | Guard setup — experimental |
Review Slips alert data | Group supplied JSONL alerts into review cases, inspect simulated decisions and export a signed evidence bundle. | SOC walkthrough — reference implementation |
Add evidence review to a delivery workflow | Integrate verification into a GitHub Action or your own agent workflow, then keep human acceptance as a separate step. |
OWP does not automatically watch every agent or tool on your computer. Fleet permissions, WorkBuddy integration and live SOC remediation are not part of the verified release paths above.
Related MCP server: Agent Receipts
Get started with MCP
You need Python 3.10+, uv, and an MCP-capable client. Add this local stdio server to the client's MCP configuration:
{
"mcpServers": {
"openworkproof": {
"command": "uvx",
"args": ["openworkproof==1.6.1"]
}
}
}This is the common mcpServers configuration shape. Some clients use a different file or top-level key; see the client-specific setup. uvx downloads the package and dependencies when needed. OWP does not require an OWP cloud account or API key.
After the client connects, it should discover 32 tools. Try this first with the included synthetic data:
Use
owp_soc_sample_alertsto get the OpenWorkProof sample. Pass itsalerts_jsonltoowp_soc_analyze_slips, then pass the returned bundle toowp_soc_verify_bundle. Explain the cases and simulated actions. Do not operate a real host. If independently obtained trusted keys and checkpoints are missing, keep the verification result UNKNOWN and explain what is missing.
You should receive a case summary and evidence bundle. UNKNOWN is expected without independent trust information; it does not mean installation failed. This sample uses a deterministic investigation stub and simulated actions. It does not test detection accuracy or remediate a host.
To use your own alerts, follow the SOC guide. Alert content is returned to your MCP client and can enter its model conversation, so use data you are authorized to share with that client.
Optional: install the SOC Review Skill
The Skill gives your agent a repeatable procedure for using those tools. With Node.js/npm available:
npx skills add dengyier/OpenWorkProof --skill openworkproof-soc-reviewChoose your agent and install scope in the installer. Read the instructions first, or use the Skill ZIP with the 2026-11-09 credit revision.
The Skill does not configure MCP or enable Guard. Configure the MCP server above separately. Your client's usual model charges still apply.
Try Codex preview protection
Experimental, separately enabled. Guard is for supported local text-editing tasks in a designated preview directory. Use a new test project or a copy you own first.
Read the English first-run guide before enabling. It includes a complete test project, scope review, expected checks and uninstall steps. The approval prompts are currently in Chinese, while the default output is JSON. Review the scope yourself and enter yes only when you understand it; do not use a trust shortcut to skip that review.
enable runs a live probe and run uses your existing Codex login and model allowance. Continue only after doctor confirms the required setup, current review and live probe. Ordinary Codex launches are not automatically protected.
The recorded live-host baseline covers macOS, Python 3.12 and Codex CLI 0.147.0, from the Guard implementation introduced in 1.5.0. Check compatibility and rerun the probe in your own environment. Builds, deployment, online search, personal MCP/plugins and whole-computer isolation are outside this launch path. See scope and limitations.
Install the CLI package, then run these commands from your target Git project after creating a preview directory and a text file such as preview/index.html:
uv tool install --python 3.12 'openworkproof==1.6.1'
owp-guard install --project "$(pwd -P)" --allow-dir preview
owp-guard enable --project "$(pwd -P)"
owp-guard doctor --project "$(pwd -P)"
owp-guard run --project "$(pwd -P)" --prompt "Update the heading in preview/index.html. Keep the change in preview; do not publish."Check an evidence bundle from the command line
Already have an OWP bundle? Install the package as above, then use the command matching its format:
# Core Surface Bundle
owp surface-verify /path/to/surface-bundle
# Core Acceptance Bundle
owp acceptance-bundle-verify /path/to/acceptance-bundle
# SOC reference output; missing external trust intentionally returns UNKNOWN/nonzero
owp-aisoc verify /path/to/soc-reviewThese formats and their trust requirements differ. Follow the Core verification guide or SOC trust guide. Obtain expected public keys and checkpoints through a separate trusted route; copying them from an untrusted bundle does not establish who you are trusting.
What a result means
Verification checks evidence under the stated scope and trust inputs. It does not establish that a business outcome is correct. Core
VERIFIEDand bindingBOUNDdo not mean customer acceptance or payment.Human acceptance is separate. The intended reviewer keeps the decision to accept or reject a delivery. See the acceptance workflow.
SOC results are reference results. Investigation is a deterministic stub; actions and acceptance statuses are simulated. There is no live SOC deployment, production detection benchmark, OS isolation or HSM in this integration.
Signing tools can handle sensitive keys. Generic key-generation/signing tools can return or accept private-key material. Keep production signing keys out of model conversations and use an externally controlled signing workflow.
OWP is open source. The current package and Skill are free to use under their licenses; your model provider and infrastructure may charge separately. Public listings, automated scores and engineering tests are not customer adoption or a security certification.
Documentation
Need | Documentation |
Connect MCP and understand the 32 tools | |
Review Slips alerts and install the Skill | |
Set up a guarded Codex task | |
Build an integration with signed task scope and human authority | |
Understand protocol mechanics and historical validation | |
Report a reproducible problem or contribute |
Community and contributing
The card is a live directory badge; follow it for the scanned version and scope. Find OWP on MCPRush, Glama, and AgentSkills.in.
The community record preserves directory contributions, external research, integration experiments and technical writing, with their original links and checked dates. These records have different scopes and do not imply endorsement or production adoption.
Contribute a reproducible bug report, an integration, or a documentation improvement. Please report security issues privately through GitHub Security Advisories.
License
OWP Core: Apache-2.0. Tim Shing(成孝添), GitHub alphatin123, contributed the aisoc component under MIT. The combined package metadata uses Apache-2.0 AND MIT.
Available Tools
32 toolsowp_analyze_repoA
Analyse a repository's structure: file count, languages, dependencies.
Args:
source: Local path or remote Git URL to the repository.
clone_enabled: If True and source is a URL, clone it to a
temporary directory first (default True).
Returns: Repository analysis including file entries, language stats, line counts, and dependency information.
Inputs: source (required), clone_enabled. Returns: object
owp_build_delivery_packageC
Export a public, diagnostic, or customer-private delivery package.
Inputs: ledger (required), output (required), privacy_view (required). Returns: object
owp_compute_digestA
Compute the canonical SHA-256 digest of an OpenWorkProof payload.
The digest is computed over the RFC 8785 JCS canonicalisation of
{"domain": "openworkproof/<object_type>/v0.1", "payload": <payload>}
(with digest and signature fields stripped from the payload
at the top level).
Args: object_type: One of the allowed canonical domains. payload: JSON string of the payload object.
Inputs: payload (required), object_type (required). Returns: object
owp_compute_key_idA
Compute the OpenWorkProof key_id for a given Ed25519 public key.
Args: public_key_b64url: The raw public key bytes encoded as unpadded base64url (43 characters).
Inputs: public_key_b64url (required). Returns: object
owp_control_observation_validateA
Assess one v0.5 control observation set against its contracts.
This read-only tool never signs, commits, accepts, or settles anything.
Signer authority is reported as not_checked, never as authorized.
Inputs: payload_json (required). Returns: object
owp_explain_binding_decisionB
Verify and explain one BindingDecision (read-only).
Inputs: payload_json (required). Returns: object
owp_generate_keypairA
Generate an Ed25519 signing key pair; this tool takes no input parameters.
Returns the private key as a hex string and the public key as
base64url, along with the derived key_id.
Inputs: no named parameters. Returns: object
owp_get_binding_statusC
Read the current binding decision head from a ledger (read-only).
Inputs: ledger_path (required). Returns: object
owp_get_decisionB
Prepare, but do not sign or commit, a versioned decision draft.
Inputs: ledger (required), request_json (required). Returns: object
owp_get_schemaA
Get the authoritative JSON Schema for an OpenWorkProof object type.
Args:
object_type: One of work-order, capability-grant,
action-receipt, acceptance-receipt,
acceptance-rejection-receipt.
version: Schema version (default 0.1).
Inputs: object_type (required), version. Returns: object
owp_get_schema_digestA
Get the frozen SHA-256 digest of an authoritative JSON Schema.
This digest is a protocol-review anchor: changing it requires an explicit version-bump decision.
Args:
object_type: One of the five schema-bearing object types.
version: Schema version (default 0.1).
Inputs: object_type (required), version. Returns: object
owp_get_settlement_readinessB
Derive readiness only; this does not prove payment or settlement.
Inputs: ledger (required). Returns: object
owp_integrity_observation_validateA
Assess one v0.5 population observation set against its contracts.
This read-only tool never signs, commits, accepts, or settles anything.
Signer authority is reported as not_checked, never as authorized.
Inputs: payload_json (required). Returns: object
owp_list_domainsA
List canonical OpenWorkProof domains; this tool takes no input parameters.
Returns:
A dict with canonical_domains (all 9) and signed_domains
(8 — excludes sidecar-event which cannot be signed).
Inputs: no named parameters. Returns: object
owp_render_surface_reportC
Return the verified derived report, never payment or acceptance.
Inputs: package_path (required). Returns: object
owp_repo_readB
Forward a repo-read execution to the ledger coordinator.
Args: ledger: Path to the SQLite ledger file. payload: JSON string of the repo-read execution payload.
Inputs: ledger (required), payload (required). Returns: object
owp_run_testsB
Run tests, write an execution receipt to the ledger, and save test evidence files.
The ledger coordinator checks the signed request before execution. The configured executor writes execution artifacts under the runtime and evidence roots supplied in the payload.
Args:
ledger: SQLite ledger file that receives the execution receipt.
payload: JSON string of the run-tests execution payload (must
include the signed AgentRequest and typed arguments).
Inputs: ledger (required), payload (required). Returns: object
owp_run_verificationA
Write verification results or signed decisions to the ledger, or prepare a draft.
operation must be commit_arm, prepare_decision, or
commit_decision. commit_arm appends a validated verification
result to the SQLite ledger. commit_decision appends the supplied
signed verification decision. prepare_decision returns an unsigned
draft without committing that decision. The tool never retries an
indeterminate commit.
Inputs: ledger (required), payload (required), operation (required). Returns: object
owp_scope_compareB
Compare a signed v0.3 scope with one verifier observation.
Inputs: scope_json (required), observed_scope_json (required). Returns: object
owp_scope_validateA
Intrinsically validate a v0.3 scope without checking authority.
This read-only tool does not accept a ledger, private key, or signature instruction. Full Manager authority is checked only by the non-MCP commit boundary.
Inputs: scope_json (required). Returns: object
owp_sign_payloadA
Sign a canonical OpenWorkProof payload with an Ed25519 private key.
Args:
object_type: One of the allowed signed domains (call
owp_list_domains to see the list).
payload: JSON string of the unsigned payload object.
private_key_hex: The Ed25519 private key as a hex string.
Returns:
The signed payload dict with signature_alg, signer_key_id,
digest, and signature fields added.
Inputs: payload (required), object_type (required), private_key_hex (required). Returns: object
owp_soc_analyze_slipsARead-only
Review supplied Slips JSONL locally and return cases plus a signed evidence bundle.
Accepts up to 128 KiB, 500 lines and 50 source windows. Uses a synthetic-trained model, a deterministic investigation stub and in-memory simulated actions. Does not control real hosts, push to AiSOC or upload telemetry. Returned keys and checkpoints are self-supplied evidence, not an independent trust anchor.
Inputs: alerts_jsonl (required), tenant. Returns: object
owp_soc_sample_alertsARead-only
Return synthetic Slips JSONL and scope for the reference SOC workflow.
No host scanning, external connections or real remediation is performed. This does not activate Guard or protect other MCP tool calls.
Inputs: no named parameters. Returns: object
owp_soc_verify_bundleARead-only
Verify a reference SOC bundle using separately acquired expected trust pins.
Inputs are JSON strings, each at most 8 MiB. Omitted external keys or checkpoint pins yield UNKNOWN, not trusted acceptance. Copying pins from the same bundle only tests consistency; caller-supplied pins do not prove organizational independence. A FAILED/UNKNOWN status is not acceptance. No filesystem writes.
Inputs: bundle_json (required), trusted_keys_json, trusted_checkpoints_json. Returns: object
owp_statusA
Replay an OpenWorkProof ledger and return its authoritative state.
Args: ledger: Path to the SQLite ledger file.
Inputs: ledger (required). Returns: object
owp_validate_action_binding_manifestB
Validate one signed ActionBindingManifest (read-only, no authority).
Inputs: payload_json (required). Returns: object
owp_validate_judgment_commitmentA
Validate one signed JudgmentCommitment (read-only, no authority).
Without a ledger or trusted key context the authority is reported as
not_checked; this tool never signs and never commits.
Inputs: payload_json (required). Returns: object
owp_validate_profileC
Validate a signed Evidence Lifecycle v0.2, v0.3, or v0.5 profile.
Inputs: profile_json (required). Returns: object
owp_verify_nested_claimA
Verify a nested claim (AgentRequest or HumanDecision) against a WorkOrder.
This checks that:
The WorkOrder identity bindings are valid.
The claim's
work_order_digestmatches the WorkOrder.The claim's actor is bound to a matching key in the WorkOrder.
The claim's Ed25519 signature is valid.
Args: claim_json: JSON string of the AgentRequest or HumanDecision. work_order_json: JSON string of the parent WorkOrder.
Returns:
{"valid": true} if the nested claim passes all checks.
Inputs: claim_json (required), work_order_json (required). Returns: object
owp_verify_signatureA
Verify an Ed25519 signature on a signed OpenWorkProof payload.
Args:
object_type: The canonical domain of the object.
signed_payload: JSON string of the signed payload (must
contain signature_alg, signer_key_id, digest,
and signature).
public_key_b64url: The signer's public key as base64url.
Returns:
{"valid": true} if the signature is valid, {"valid": false}
otherwise.
Inputs: object_type (required), signed_payload (required), public_key_b64url (required). Returns: object
owp_verify_surface_bundleB
Verify and replay one surface bundle without writing or signing.
Inputs: package_path (required). Returns: object
owp_verify_work_orderA
Verify a WorkOrder's identity bindings and maintainer signature.
Args: work_order_json: JSON string of the WorkOrder object.
Returns:
{"valid": true} if all six key bindings are well-formed,
distinct, and the maintainer's signature is valid.
Inputs: work_order_json (required). Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 32 tools
Most tools are distinguishable by object type plus action (sign vs. verify vs. validate a given artifact), but several boundaries blur: `owp_get_decision` overlaps with `owp_run_verification`'s prepare_decision mode, and `owp_validate_profile` sits awkwardly beside the scope/observation validators. The verification family (`verify_signature`, `verify_work_order`, `verify_nested_claim`, `verify_surface_bundle`, `soc_verify_bundle`) is large and requires careful description reading to pick correctly.
Nearly all names use an `owp_` prefix with consistent snake_case verb_noun structure (validate_profile, verify_signature, compute_digest, get_schema). Minor deviations exist: `owp_status` and `owp_repo_read` are noun-first, and the SOC tools insert a namespace segment (`owp_soc_*`), but the pattern remains readable and predictable.
At 32 tools this is well past the range where an agent can hold the surface in mind, and the breadth comes partly from many narrow, near-duplicate read-only validators (profile, scope, integrity observation, control observation, judgment commitment, action binding manifest). The protocol is genuinely broad, but the set could be consolidated considerably.
The surface covers signing primitives, key management, canonical digests, schema retrieval, ledger replay/status, decision drafting and commit, delivery packaging, settlement readiness, plus a full SOC and repo-analysis workflow. Gaps remain on the authoring side: WorkOrders, capability grants, and acceptance receipts can be verified or schema-fetched but have no dedicated creation/commit tool, forcing reliance on generic `sign_payload`.
Maintenance
Related MCP Connectors
Tamper-evident proof creation and verification for AI agents via MCP, A2A, and REST.
Verifiable execution records for AI agents: signed receipts, tamper-evident archives and portable evidence bundles, exposed via MCP and REST, with paid verify and anchor calls settled over x402.
Public MCP for agent verification, work discovery and governed interoperability.
Agent-native MCP for governed commerce, x402 payments, paid capabilities, and verifiable receipts.
Related MCP Servers
- AlicenseCqualityBmaintenanceCryptographic identity and trust protocol for AI agents. 38 MCP tools across 8 protocol layers: Ed25519 identity, delegation chains, values compliance, signed communication, policy engine, task coordination, cross-layer integration, and agentic commerce. 264 tests passing.154262 npm4Apache 2.0
- AlicenseAqualityBmaintenanceCryptographic accountability for AI agents. Ed25519-signed receipts for every MCP tool call. Constraints, chains, AI judgment, invoicing, and local dashboard included.241MIT
- AlicenseNot gradedqualityCmaintenanceTamper-evident audit logging for AI agents. Append-only, hash-chained, optionally Ed25519-signed log. The MCP server lets an agent keep and verify a record of what it actually did.8MIT
- AlicenseNot gradedqualityCmaintenanceA sovereign, MIT-licensed MCP server for professional-service workflows, providing offline-capable, Ed25519-signed tools for autonomous agents and human developers.MIT