sswp-mcp
The SSWP MCP (Sovereign Software Witness Protocol) server provides deterministic software attestation, supply-chain risk analysis, fleet registry management, and tamper-proof auditing for software repositories. It operates as an MCP server, allowing AI agents (Claude, VS Code Copilot, etc.) to interact with its tools natively.
sswp_witness— Full cryptographic witness on a single repository: scans the dependency graph, runs a 5-gate pipeline (GIT_INTEGRITY, LOCKFILE, DETERMINISTIC_BUILD, TEST_PASS, LINT), adversarially probes dependencies for typosquatting/version anomalies/missing integrity hashes, and seals results into a self-verifying SHA-256-backed.sswp.jsonattestation.sswp_verify— Recompute the SHA-256 signature of an existing.sswp.jsonto detect tampering.sswp_bulk_witness— Run the full witness pipeline sequentially across multiple repositories in one call — ideal for nightly audits or pre-release sweeps.sswp_check_repo— Lightweight health check (.git,package.json,package-lock.json) without a full witness run, useful as a fast CI pre-check.sswp_analyze_deps— Analyze a list of npm packages for supply-chain risk (typosquatting, version anomalies, missing metadata integrity hashes), with optional AI-powered deep reasoning.sswp_registry_health— Fleet health dashboard showing all witnessed nodes with status, last run time, overall risk score, and adversarial risk score, ordered by risk.sswp_ledger— Query the append-only SHA-256 hash-chained audit ledger recording every witness run, gate vote, and probe result for compliance or incident investigation.sswp_node_search— Full-text search (FTS5) across the fleet registry by repository name, tag, or description keyword.
Integrates with Ollama to run Kimi K2 reasoning for deeper adversarial probing of dependencies.
SSWP MCP
SSWP records what configured software checks actually establish. Its MCP and CLI entry points share one implementation and a persistent SQLite registry.
Version 2.1 fixes false Git PASS results, missing execution timeouts, source/bundle divergence, incomplete nested hashes and inconsistent ledger timestamps. Witness execution requires a short-lived operator-policy receipt from Omega Brain. Caller-supplied approval strings are no longer accepted.
Use Node 22 or newer and pnpm 11. Install with pnpm install --frozen-lockfile, run pnpm build, then start node dist/sswp.cjs. pnpm check checks TypeScript. Build receipts contain source, lockfile and artifact hashes. Native SQLite and esbuild install scripts are explicitly allowed by the workspace configuration.
Configure an absolute SSWP_DB path and the same VERITAS_SHARED_DIR used by Brain and Stenographer. Add the server command to persistent user MCP configuration. sswp_check_repo is read-only; sswp_witness runs only explicitly configured project commands, checks the receipt, writes an attestation and durably queues an event for Brain.
PASS, FAIL, ERROR, INCONCLUSIVE and NOT_APPLICABLE remain distinct. BUILD_SUCCESS and REPRODUCIBLE_BUILD are separate gates. Reproducibility requires nonempty matching outputs from two fresh source copies. sswp_verify covers nested fields in v2 attestations and reports older formats as LEGACY_UNVERIFIED; a content hash does not authenticate an author.
sswp_registry_health reports the latest run per node, timestamps, current path availability, pending materializations and delivery status. Registry writes, gate rows, ledger events and outbox jobs commit together. Committed attestations can be materialized and delivered again after interruption. Historical records remain historical.
See current operating contract for exact permission boundaries, missing-evidence behavior, dependency coverage and recovery. Tests use local synthetic projects. pnpm test additionally exercises Python-to-Node receipts when an Omega Brain checkout and .venv are present as siblings, as arranged by CI.
Available Tools
8 toolssswp_analyze_depsAIdempotent
Analyze a list of dependencies for supply-chain risk using Kimi K2 reasoning. Provide an array of {name, version} objects for any npm packages you want evaluated. The tool performs four analysis passes: typosquatting detection (matching names against known suspicious patterns like left-pad, event-stream), version anomaly scanning (flagging unpinned ranges like *, >=, ^0), metadata integrity checks (CRITICAL if a dependency lacks an integrity hash), and optional Kimi K2 deep reasoning (requires OLLAMA_CLOUD_API_KEY — returns INCONCLUSIVE without it). Returns a JSON object with per-probe results, overall risk score (0-1), and suspicious package counts. Use this for targeted supply-chain analysis on critical dependency trees. For generating full attestations that include probing, use sswp_witness.
Inputs: packages (required).
sswp_bulk_witnessAIdempotent
Run deterministic attestation on multiple repositories sequentially. For each repo path provided, runs the full SSWP witness pipeline (scan, 5-gate test, adversarial probe, SHA-256 seal) and auto-saves the .sswp.json attestation to the fleet registry. Reports per-repo PASS/FAIL status with risk percentages and a final summary of passed, failed, and skipped counts. Missing repos are skipped by default. Use this for nightly fleet audits, pre-release sweeps across the ecosystem, or any batch witnessing operation. For a single repo, prefer sswp_witness.
Inputs: repoPaths (required), skipMissing.
sswp_check_repoARead-onlyIdempotent
Perform a lightweight repo health check without running the full witness pipeline. Verifies four conditions: the directory exists on disk, a .git directory is present (indicating a git repository), a package-lock.json exists (indicating locked dependencies), and a package.json exists (indicating a valid Node.js project). Returns a status line for each condition and an overall READY/NOT READY verdict. Use this as a fast pre-check in CI pipelines or before calling sswp_witness to ensure the repo is in a valid state. Does not seal an attestation or modify the registry.
Inputs: repoPath (required).
sswp_ledgerARead-onlyIdempotent
Query the tamper-proof SSWP audit ledger, an append-only SHA-256 hash chain that records every witness run, gate vote, and probe result. Returns a formatted table showing ledger entries with their sequence ID, event type (WITNESS, BULK_WITNESS), hash, and timestamp. Optionally filter by event type to narrow results. The ledger chain is cryptographically verifiable — any altered or removed entry breaks the chain. Use this for audit trail review, compliance reporting, or incident investigation. For a quick fleet overview, use sswp_registry_health instead.
Inputs: limit, eventType.
sswp_node_searchARead-onlyIdempotent
Search the SSWP fleet registry using full-text search (FTS5) across node names, tags, and descriptions. Matches partial keywords and ranks results by relevance. Returns a formatted table showing matching nodes with their name, node type, status, and repository path. Use this to find specific projects in the ecosystem registry by name fragment, technology tag, or description keyword. For a full sorted health board of all nodes, use sswp_registry_health instead.
Inputs: query (required), limit.
sswp_registry_healthARead-onlyIdempotent
Display the full fleet health board from the SSWP SQLite registry. Returns a formatted table showing every witnessed node with its name, status (active/deprecated/archived), last witness run timestamp, overall risk score (as percentage), and adversarial risk score (as percentage). Results are ordered by risk descending (most risky nodes first). Use this for an ecosystem-wide dashboard view of attestation status. For searching specific nodes by name, tag, or description, use sswp_node_search. For querying the audit ledger directly, use sswp_ledger.
Inputs: limit.
sswp_verifyARead-onlyIdempotent
Verify the SHA-256 cryptographic signature of an existing .sswp.json attestation file. Recomputes the hash over the entire attestation payload (sorted keys, excluding the signature field) and compares it against the stored signature. Returns VALID ATTESTATION if the file is intact and unmodified, or SIGNATURE MISMATCH if the file was altered after sealing. Use this to audit an attestation you received from someone else, or to confirm a repo's attestation still matches the file on disk. For generating new attestations, use sswp_witness; for quick repo readiness checks without sealing, use sswp_check_repo.
Inputs: filePath (required).
sswp_witnessAIdempotent
Witness a software project with deterministic attestation. Scans the full dependency graph (every node_modules package with resolved path, integrity hash, and risk score), runs a 5-gate pipeline (GIT_INTEGRITY, LOCKFILE, DETERMINISTIC_BUILD, TEST_PASS, LINT), adversarially probes every dependency for typosquatting, version anomalies, and missing integrity hashes, then produces a self-verifying .sswp.json attestation sealed with SHA-256. Auto-saves the attestation to the SQLite fleet registry and appends an entry to the tamper-proof audit ledger. This is the primary attestation tool — use it when you need a full cryptographic witness of a single repo's state. For multiple repos, use sswp_bulk_witness instead.
Inputs: repoPath (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 8 tools
Each tool has a clearly distinct purpose: single attestation, bulk attestation, verification, dependency analysis, pre-flight checking, registry health, ledger audit, and node search. Even the similar single vs. bulk witness tools are explicitly differentiated by scope.
All tools share the sswp_ prefix and snake_case, but base names mix conventions: verb-only (witness, verify), verb_noun (analyze_deps, check_repo), and noun forms (ledger, registry_health, node_search). The domain prefix keeps it readable, but there's no single verb_noun pattern.
8 tools is well-scoped for a supply-chain attestation workflow: single and bulk witnessing, verification, pre-check, dependency analysis, and three query/audit surfaces. No redundancy or bloat.
The lifecycle is well covered: witness, bulk witness, verify, pre-check, dependency analysis, registry health, ledger audit, and search. Minor gaps exist around registry lifecycle management (e.g., no explicit tool to update node status or remove a node), but core workflows are covered.
Maintenance
Related MCP Connectors
Governed MCP: agent audit, provenance, deterministic checks, and receipt-backed FragGate execution.
Attestation infrastructure for the agentic economy: signed, independently verifiable verdicts.
Control plane for autonomous software labor. Agents claim objectives over MCP with audit trail.
Pre-execution governance for AI agents. Deterministic PASS/FAIL/REVIEW verdicts, replayable proof.
Related MCP Servers
- AlicenseNot gradedqualityFmaintenanceGovernance kernel for AI agents — policy enforcement, code safety verification, multi-model hallucination detection (CMVK), trust attestation (IATP), and immutable audit trails. Works with Claude Desktop, Cursor, and any MCP client.74MIT
- AlicenseAqualityBmaintenanceCryptographic accountability for AI agents. Ed25519-signed receipts for every MCP tool call. Constraints, chains, AI judgment, invoicing, and local dashboard included.241MIT
- AlicenseAqualityCmaintenanceAI agent provenance, trust, and auditability layer. VERITAS multi-gate scoring, Cortex approval gates, S.E.A.L. hash-chain audit ledger, and semantic RAG with cryptographic provenance tracking for every decision an agent makes.275MIT
- AlicenseBqualityBmaintenanceA governance proxy for AI tools — every MCP/agent tool call is policy-gated, secret-redacted, and written to a hash-chained, offline-verifiable audit trail.13MIT