Documentation · 11 / 12
Agent access
The public functions as tools for AI agents. No API key.
Four ways in
- MCP server at /mcp (Streamable HTTP)
- A2A agent at /a2a with an agent card at /.well-known/agent-card.json
- WebMCP on the page /agents/ for agents inside the browser
- JSON API under /agents/api, described by /agents/openapi.json
What agents can do
Search projects and tenders, read service areas, countries and provider coverage, look up vulnerabilities, read guides, run the CONTENTUM Check, test the NIS2 scope, send a request for a company or a provider and ask for its status.
| Tool | What it does | Kind |
|---|---|---|
search_projects | Open cybersecurity and IT infrastructure projects and public tenders by service, country and words. | reads |
get_project | One project or tender by id (CT-0001): services, country, signal, regulations, technologies, deadline. | reads |
list_services | The 12 service areas with their open project counts. | reads |
get_service | One service area: description, typical tasks, demand signal, open projects, provider count. | reads |
list_countries | Countries with counts of company projects, tenders and listed providers. | reads |
get_country | One country: open projects by service, tenders, providers by service, latest projects. | reads |
get_provider_coverage | How many listed providers offer a service in a country (counts only). | reads |
get_market_stats | Totals: open projects, tenders, countries, demand per service and regulation, data date. | reads |
search_vulnerabilities | CVE records by vendor or year, or the actively exploited ones (CISA KEV). | reads |
get_vulnerability | One CVE: description, CVSS, weaknesses, references and whether it is actively exploited. | reads |
list_guides | Buyer guides (NIS2, penetration testing, ISO 27001, OWASP …) in a given language. | reads |
get_guide | Full text of one guide with sections, questions and sources. | reads |
run_domain_check | Passive security check of a domain: TLS, security headers, SPF, DMARC, security.txt. Returns score and report link. | sends data |
check_nis2_scope | Whether an organisation is probably covered by NIS2, from sector and size. | reads |
get_platform_info | What CONTENTUM does, what a request needs, contacts, languages and the agent endpoints. | reads |
submit_project_request | A company describes the work it needs. Returns a request id; the team replies by e-mail. | sends data |
register_provider_interest | A service provider asks to be listed or registers interest in a project. Returns an id. | sends data |
get_request_status | Status of a request (id + e-mail). | reads |
Limits and privacy
Reading tools are open, with a limit per address. Tools that send data have a tighter limit. Agents see what the public pages show: no company names, and providers as counts only.
The domain map and the Deep Scan are not open to anonymous agents, because they need the permission of the owner.
Machine-readable descriptions
llms.txt and /agents/llms.txt describe the site and the tools for language models. The OpenAPI file, the agent card and the MCP server card describe the interfaces. This documentation is available as one Markdown file per language at /docs/docs.md.
Use cases
An agent finds a provider for its principal
- Call search_projects or get_provider_coverage to see the market.
- Call submit_project_request with company, e-mail and description.
- Keep the request number.
- Call get_request_status later and report back.
An agent answers a security question
- Call get_vulnerability for the CVE in question.
- Call list_guides and get_guide for the background.
- Call run_domain_check for a domain the principal is authorised to check.
- Give the answer with the links from the results.
User stories
- As an AI agent, I want to send a request without a form or captcha, so that I can finish the task for my principal.
- As a developer, I want an MCP endpoint without registration, so that I can connect my assistant in a minute.
- As the operator, I want agents to see only public data, so that no company name leaks.
Updated: