Compliance AIops

Compliance AIops

Audit availability: Real writes require a committed pending audit row and stop before contacting the platform if storage is unavailable. Reads and previews may continue with a warning; new audit rows store outcome summaries rather than full responses.

Disclaimer: Community-maintained open-source project. Not affiliated with, endorsed by, or sponsored by any framework body or GRC vendor. HIPAA, PCI-DSS, SOC 2, GDPR and OSCAL are referenced descriptively; the frameworks and trademarks belong to their owners. MIT licensed.

Governed compliance-evidence tooling for AI-agent infrastructure ops. It reads the audit trails your governed AIops agents already write — the local ~/. -aiops/audit.db SQLite trails, all sharing one audit_log schema — and turns that activity into framework-mapped, hash-chain-sealed compliance evidence. It never scans your infrastructure and never replaces a GRC platform: it converts the trails you already produce into auditor-ready, tamper-evident evidence bundles.

Unlike the other tools in the AIops-tools line it is not a platform wrapper: no external API, no network, no platform credentials. Its only inputs are those on-disk audit databases, read read-only. That also makes it the easiest-to-self-test tool in the line — fully offline and deterministic.

Evidence, not certification. Fully offline; the source audit.db

files remain the system of record. OSCAL export is a documented v0.2 roadmap

item (v0.1 emits JSON + Markdown + CSV shaped to ease a future OSCAL

Assessment-Results adapter).

Key features

• Framework mapping with honest evidence-strength — audit events map to HIPAA §164.312 / PCI-DSS v4.0 / SOC 2 TSC / GDPR controls. Audit trails prove *operating effectiveness* strongly but *control design / configuration* only partially, and each control is labelled strong or partial. gap_analysis says so per control, with the caveat and a remediation hint.
• Hash-chain-sealed evidence bundles — SHA-256 over ordered records (hash = SHA-256(prev_hash ‖ canonical_json(record)), genesis prev = 64 zeros). The chainHead is reproducible for the same (framework, period, sources). verify_bundle catches tampering; verify_source_chain detects row-id gaps / deletions in a source trail. An optional HMAC signature seals a bundle under a stored signing key.
• Zero-network, read-only — no credentials, no outbound calls, no mutation of the source trails. Bundles are the only thing written, under ~/.compliance-aiops/bundles/.
• Deterministic, test-verified integrity — the integrity claims are themselves covered by tests: synthetic audit DBs are built through the real governance-harness AuditEngine, a golden reproducible chainHead is asserted, and tamper tests confirm detection. No live infrastructure needed.

What this tool does, and does not, decide

It reads your audit trails and writes evidence bundles — and records every operation. It does not decide whether producing or signing a bundle is allowed: that is the agent's judgement, or the filesystem permissions of the account it runs as. The source audit.db files are opened strictly read-only regardless, and the only thing ever written is a bundle under ~/.compliance-aiops/bundles/.

So there is no read-only switch, no policy file, no approval gate to configure. The one thing the tool guarantees is that nothing is silent: every call, over MCP and over the CLI alike, lands an audit row in ~/.compliance-aiops/audit.db.

Each tool declares a risk_level, kept in agreement with its [READ]/[WRITE]

documentation tag by a test, and carried into the audit row as a descriptive

tier — so a reviewer can see at a glance what a row was. It is a label, not a

gate.

Tools (16 MCP tools)

Read / analysis (13)

Write / artifact (3 — no external mutation)

The CLI exposes a convenience subset; the full 19-tool surface is available over MCP.

Frameworks & controls

Install

uv tool install compliance-aiops      # or: pipx install compliance-aiops

Quick start

As a Claude Code plugin

One install gives an agent both the skill and the MCP server:

/plugin marketplace add AIops-tools/marketplace

/plugin install compliance-aiops@aiops-tools

The MCP server is fetched with uv and pinned to the package version this plugin declares, so an audit row can be traced back to the code that wrote it. Credentials are still configured with compliance-aiops init — see below.

As an OpenClaw plugin

The same bundle is published on ClawHub, where one install delivers the skill and its MCP server together:

openclaw plugins install clawhub:@zw008/compliance-aiops

openclaw skills info compliance-aiops # expect: Visible to model: yes

Restart the OpenClaw gateway afterwards so it loads the plugin. The MCP server is fetched with uv, pinned to this exact release, so uvx has to be on PATH — without it the skill still installs but reports Visible to model: no. Credentials are configured exactly as below.

As a CLI or standalone MCP server

compliance-aiops init                 # discover sibling ~/.*-aiops/audit.db, set org name, optional signing key

compliance-aiops doctor # which sibling audit DBs are present/readable compliance-aiops overview # audit sources + per-framework covered/total compliance-aiops report coverage soc2 # per-control SOC 2 coverage compliance-aiops bundle generate soc2 # sealed evidence bundle → ~/.compliance-aiops/bundles/ compliance-aiops bundle verify # re-verify the chain + seal (+ signature)

Run as an MCP server (stdio):

export COMPLIANCE_AIOPS_MASTER_PASSWORD=...   # only needed to unlock a signing key

compliance-aiops mcp # or: compliance-aiops-mcp

Where that password then lives: an exported variable is readable by

every process this shell starts and is recorded by shell history. On a

shared or long-lived host, prefer the interactive prompt, or inject it from

a secret manager for the life of the one command that needs it.

定期封存 (scheduled sealing)

Evidence bundles are most useful when sealed on a cadence (e.g. weekly), so each period has a tamper-evident anchor. This tool ships no daemon — instead bundle schedule prints a ready-to-paste cron line plus the exact non-interactive command, and writes nothing:

compliance-aiops bundle schedule soc2 --cron "0 2 * * 1" --period 7d --sign

It returns a cronLine you paste into crontab -e, for example:

0 2 * * 1 compliance-aiops bundle generate soc2 --period 7d --sign

• --period (also available on bundle generate) is a convenience relative window — 7d, 24h, 2w, or last-7-days — resolved to a since/until pair ending "now", so each scheduled run seals the trailing window.
• Export COMPLIANCE_AIOPS_MASTER_PASSWORD in the cron job's environment so a stored signing key unlocks non-interactively. Do not inline the real password in the crontab file — reference it from a protected env file.

Integrity & honest limits

• Tamper-EVIDENT, not tamper-PROOF. The hash chain and optional signature let an auditor *detect* alteration; they do not prevent it. The source audit.db files remain the system of record — record the chainHead out-of-band if you need an independent anchor.
• Operating effectiveness vs. design. An audit trail strongly evidences that a control *ran* (samples, approvals, denials) but only partially evidences that a control is *designed / configured* correctly (e.g. MFA required, least-privilege roles). Every control carries a strong / partial label and gap_analysis surfaces the caveat rather than overclaiming.

Supported scope & limitations

• Evidence, not certification. This produces auditor-ready evidence bundles; it does not issue attestations, opinions, or certifications.
• In scope: the six frameworks above, over the audit_log trails written by governed AIops tools discovered via ~/.*-aiops/audit.db.
• Not in scope: it does not scan infrastructure, connect to any platform, or replace a GRC platform. For platform operations use the other AIops-tools.
• OSCAL export is v0.2. v0.1 emits JSON + Markdown + CSV.
• Interfaces may change before v1.0.
• Verification: the integrity claims are covered by deterministic offline tests; see docs/VERIFICATION.md for the reproducible run.

Missing a capability?

Want another framework, control mapping, export format (OSCAL, CSV shape), or a verification you don't see here? Open an issue or a PR — contributions welcome.