Skip to main content

Documentation

Everything you need to get started with AuditClaw, from installation to advanced configuration.

Quick Start

1

Install AuditClaw

Install from ClawHub (recommended) or download the ZIP manually.

Option A: Install from ClawHub

$ openclaw skills install @mailnike/auditclaw-grc

Option B: Manual download

$ cd ~/clawd
$ curl -LO https://www.avansaber.com/apps/auditclaw/auditclaw-latest.zip
$ unzip auditclaw-latest.zip

SHA-256: 177abf39a5c6482cbfb4edcaa5d61220f5ed564b87aea9e10767abd8cdad6733

2

Initialize the database

Set up the compliance database. This creates all required tables automatically.

$ python3 scripts/init_db.py
3

Activate your first framework

Tell the bot which compliance framework you need. Just chat naturally.

You: activate soc2 framework
Bot: SOC 2 Type II activated. 43 controls loaded.
4

Start managing compliance

Ask anything. The AI understands natural language.

You: what's my compliance score?
You: add evidence "AWS Config Report" for soc2
You: show me controls that are overdue
You: generate a compliance report

Core Concepts

Frameworks

Compliance standards like SOC 2, ISO 27001, HIPAA. Each framework has pre-built controls and cross-mappings to other frameworks. Activate/deactivate at any time.

Controls

Individual compliance requirements within a framework. Track status (not_started, in_progress, complete), assign owners, set due dates, measure effectiveness.

Evidence

Proof that a control is implemented. Can be manual uploads or auto-collected from AWS/GitHub. Has validity periods and cross-framework mappings.

Integrations

Connections to cloud providers (AWS, GitHub) for automated evidence collection. Register once, collect continuously via cron schedules.

Drift Detection

Monitors your compliance over time. If something that was passing starts failing, you get an alert immediately.

Alerts

Proactive notifications for expiring evidence, declining scores, connection errors, and compliance changes. Acknowledge and resolve through chat.

Action Reference

AuditClaw has 97 built-in commands organized by category. Here are the most commonly used ones.

Status & Scoring

CommandWhat it does
statusDatabase overview: frameworks, controls, evidence counts
activate-frameworkLoad a framework and its controls
deactivate-frameworkSoft-delete a framework (preserves data)
score-historyCompliance score trends over time

Evidence Management

CommandWhat it does
add-evidenceCreate evidence record linked to controls
list-evidenceList evidence with filters
auto-map-evidenceSuggest cross-framework control links
list-evidence-gapsFind controls without evidence

Integrations & Monitoring

CommandWhat it does
add-integrationRegister a cloud provider (AWS, GitHub, Google, Okta)
integration-healthHealth dashboard for all integrations
check-driftDetect configuration regressions
compliance-calendarUpcoming deadlines from all sources
compliance-digestActivity summary (daily/weekly/monthly)

Full reference for all 97 commands available on ClawHub and GitHub.

Slash Commands

Quick-access commands for common tasks. Type these directly in your chat.

/grc-scoreQuick compliance score with trend
/grc-gapsTop priority gaps across frameworks
/grc-scanSecurity scan menu (headers/SSL/GDPR)
/grc-reportGenerate compliance report
/grc-risksRisk register summary
/grc-incidentsActive incidents overview
/grc-trustGenerate trust center page

Architecture

How AuditClaw Works
User (Chat)
|
OpenClaw Gateway
|
AuditClaw Skill (LLM Agent)
|
|
|
Compliance Engine (97 commands)
AWS Check Groups (15)
GitHub Check Groups (9)
|
Compliance Database

Skills

  • auditclaw-grc: Core GRC skill
  • auditclaw-aws: AWS evidence
  • auditclaw-github: GitHub evidence
  • auditclaw-azure: Azure evidence
  • auditclaw-gcp: GCP evidence
  • auditclaw-idp: Google/Okta

Database

  • SQLite (self-contained)
  • 97 built-in commands
  • JSON output format
  • Auto-migrations included

Testing

  • 431 automated tests
  • Edge cases and error handling
  • Python 3.10+
  • 100% command coverage

Security model: your database, your access choices

AuditClaw runs as a skill inside your OpenClaw gateway. The compliance database is a local SQLite file. Companion skills query the cloud accounts you configure and save evidence to that database. Your gateway's chat channels, tools and model provider determine how the agent processes that evidence.

Local records and exports

The database is stored at ~/.openclaw/grc/compliance.sqlite. The credential store writes provider files with owner-only file permissions, under ~/.openclaw/grc/credentials/. Treat these files and your backups as sensitive operational records.

Trust-center generation creates a local HTML file. You choose whether and where to publish it. Review evidence exports before sharing them with an assessor or customer.

Provider permissions

Choose credentials for the checks you need. AWS uses the SDK credential chain; GitHub's collector reads GITHUB_TOKEN. The credential store also records authentication methods, but that record alone does not configure a companion's runtime credentials.

Use the provider access table and test the selected integration before running a sweep. Review token permissions at the provider rather than assuming a read-only check makes the token read-only.

Operate the gateway within a trusted boundary

Keep OpenClaw updated, restrict gateway access and review enabled tools and model routing. Run openclaw security audit and follow the official OpenClaw security guidance. Separate gateways and credentials when people should not share the same access. Check the current security advisories when selecting a release.

Implementation references: credential storage, database setup and trust-center generation. Reviewed 15 October 2026. See the current skill installation reference.

AWS Integration Setup

1. Review the AWS access policy

Review the companion's included IAM policy. Give the scanning identity the permissions needed for your chosen checks. Resource checks inspect configuration; the credential-report check also requests an IAM credential report.

2. Configure credentials

The collector uses boto3.Session() and its normal credential chain. An attached role or a configured local profile can supply credentials. If you use a named profile, select it in the process that runs the collector.

export AWS_PROFILE=auditclaw

3. Register the integration

Tell AuditClaw about your AWS account.

You: add aws integration called "Production AWS"

4. Run your first sweep

Collect evidence from the 15 AWS check groups. Review any access-denied results before changing permissions.

You: run full aws evidence sweep

GitHub Integration Setup

1. Create a personal access token

Create a token limited to the repositories and organisation evidence you need. The companion documents classic repo, read:org and security_events scopes for full coverage. Classic scopes can grant write access too. Review the collector's permission requirements before granting them; a configuration-reading check does not reduce the token's authority.

2. Set the token

export GITHUB_TOKEN=ghp_your_token_here

3. Register and sweep

You: add github integration named "Main Org"
You: run full github evidence sweep

Ready to get started?

AuditClaw is free, open source, and ready for production.

AuditClaw is MIT licensed software, built by AvanSaber Inc.