compliance-intelligence

Compliance MCP Skill Example

A minimal, working example of an AI agent skill that uses the TheArtOfService Compliance MCP as its source-grounded knowledge layer.

The hosted server is listed on Smithery and in the official MCP registry as com.theartofservice/compliance-intelligence. Endpoint: https://api.theartofservice.com/mcp (streamable HTTP, anonymous access).

Two working examples: one for compliance research, one for the course catalogue. This is a reference implementation. Clone it, swap the framework name to whatever your agent needs, plug it into Claude / GPT / Copilot / your own agent runtime. Apache 2.0 licensed, no strings attached.

What this does

Given a compliance framework name (default: NIST SP 800-161), this script:

1. Connects to the compliance API at https://api.theartofservice.com
2. Pulls the framework metadata
3. Pulls all controls in the framework
4. For each control, fetches the auditor evidence requirements (categories, artefacts, common gaps, source citations)
5. Generates a structured Markdown compliance brief

The output is suitable for piping into an LLM as context, or directly handing to an auditor or risk team.

Why this pattern works

The compliance corpus behind the API is source-grounded against the published standard text and human edited, not LLM-generated. That makes it the authoritative knowledge layer for any agent that needs to reason about compliance controls.

Your agent provides the tribal knowledge layer (the organization's specific context, the tone, the workflow). The platform provides the official knowledge layer. The split is what makes the resulting brief defensible.

Stats

• 723 compliance frameworks
• 20,400+ controls
• 332,000+ cross-framework mappings
• 314,000+ courses in the catalogue, searchable without a key
• 28,586+ controls carry structured auditor evidence requirements
• Source-grounded, version-tagged, refreshed weekly

Quickstart

git clone https://github.com/GJB65/compliance-mcp-skill-example.git
cd compliance-mcp-skill-example
pip install -r requirements.txt
cp .env.example .env
# Edit .env and add your TAOS_API_KEY from https://compliance.theartofservice.com/settings
python run.py "NIST SP 800-161"

The script will print a Markdown brief to stdout. Redirect it to a file or pipe it into your downstream agent.

python run.py "ISO 27001:2022" > iso_27001_brief.md

Second example: the course catalogue

run.py answers what does this framework require. find_courses.py answers the other half: what can the user actually do about it.

python find_courses.py "soc 2 evidence collection"
python find_courses.py "first 90 days as CISO" --framework "ISO 27001"
python find_courses.py --overlap "SOC 2,ISO 27001"
python find_courses.py --frameworks

No API key needed. The four catalogue tools are free and unmetered, because they exist to help a buyer find the right course rather than to meter access to data.

The interesting one is --overlap. A plain product listing cannot answer "we are running SOC 2 and ISO 27001 at the same time, what covers both", because that is a join, not a search. An organisation in that position does not want one course per standard, it wants the overlap:

# Courses covering SOC 2,ISO 27001 together

### SOC 2 Type 2 Security controls in ISO 27001
- Price: $299.00 USD
- Covers: ISO 27001, SOC 2
- Buy: https://store.theartofservice.com/...

When nothing covers the whole combination, it says so and reports what exists per standard rather than returning an empty list.

Link buyers to the buy_url field. It carries attribution, which is how the catalogue knows an agent produced the sale.

Get an API key

1. Sign up free at compliance.theartofservice.com/register.
2. Your API key (prefix tas_) is in your account settings.
3. Free tier: 100 calls/month. Professional ($49/mo): 10,000 calls/month plus overage at $0.005/call.
4. For data licensing, white-label, or volume pricing, see the developer portal.

Using as a Claude skill

The SKILL.md file in this repo is structured as a Claude skill definition. Drop the repo into your Claude Code or Claude Desktop skills folder and Claude can invoke run.py directly with a framework name.

API endpoints used

This example hits the public REST agent API. The same data is available via the MCP endpoint at https://api.theartofservice.com/mcp if you prefer Model Context Protocol over REST.

Endpoint	Used for
GET /api/agent/frameworks/{name}	Framework metadata
GET /api/agent/frameworks/{name}/controls	All controls in the framework
GET /api/agent/controls/{code}	Full control detail including evidence_requirements
GET /api/agent/courses	Search the course catalogue (free, no key)
GET /api/agent/courses-for-frameworks	Courses covering two or more standards together
GET /api/agent/courses/{product_id}	Full detail for one course
GET /api/agent/course-frameworks	Standards covered, with course counts

Full tool catalog: compliance.theartofservice.com/developers

License

Apache 2.0. See LICENSE.

Contributions

Issues and PRs welcome. If you build a derivative skill (vendor risk, SOC 2 audit prep, framework-specific advisor, etc.) and want it linked from the example registry, open a PR with a one-line description and a repo URL.