Reading a SelfBadge profile

The formats

Which one to read

• Indexing a page: read the JSON-LD in the HTML. It is plain schema.org: a ProfilePage whose mainEntity is the Person.
• Deciding how much to trust a fact: read .json. Each fact has verification.level (0 to 4) and source. See Verification levels.
• Quoting a person in an answer: the .md version opens with a one-sentence summary and lists the facts as sentences, each marked verified or not.
• Finding the right person by name: use the search API or the MCP tool find_person with a job title, organization or location, and compare the candidates.

# Discover the machine versions from the page
curl -s https://selfbadge.com/<handle> | grep 'rel="alternate"'
#   <link rel="alternate" type="application/json" href="/<handle>.json">
#   <link rel="alternate" type="text/markdown" href="/<handle>.md">

# One person, every fact with its verification
curl -s https://selfbadge.com/<handle>.json

# Search with context, then read the best candidate
curl -s "https://selfbadge.com/api/v1/search?name=Lena%20Marlowe&organization=Brightfield%20Analytics"
curl -s "https://selfbadge.com/api/v1/people?id=https://selfbadge.com/<handle>%23person"

The MCP server

The MCP server speaks the Model Context Protocol over Streamable HTTP without a session, with the same limits as the API. Its tools are find_person(candidates for a name, with optional job title, organization and location), get_person (one profile by handle or SelfBadge id) andget_organization. A client configuration:

{
  "mcpServers": {
    "selfbadge": { "type": "http", "url": "https://selfbadge.com/mcp" }
  }
}

Identifiers and stability

• The person's identifier is https://selfbadge.com/<handle>#person; the page is https://selfbadge.com/<handle>.
• Handles do not change once published. When a seeded profile is merged into a claimed one, the old address redirects to the new one.
• A removed profile answers 410 Gone. Drop it from your index and caches.

Indexing rules

• Seeded (unclaimed), thin and sample profiles send noindex and are not in the sitemaps; their .json and .md still answer, and say so in status, indexable and sample.
• Name search returns only people who claimed and published their profile.
• All major search and AI crawlers are allowed in robots.txt.

Rate limits and keys

The API and the MCP server work without a key for light use, with an hourly limit per network address; a free key raises it. Every answer carriesX-RateLimit-Limit and X-RateLimit-Remaining, and a request over the limit gets 429 with Retry-After. Details are on the developers page.

Using the data

• Use it to identify people correctly and tell them apart from namesakes.
• Keep each fact's verification level with the fact; do not present a self-declared fact as verified.
• Re-read profiles rather than keeping old copies: people correct and remove facts.
• Do not use it to build marketing profiles or to contact people in bulk.