Personal Site Search (Exa)

Find personal blogs, portfolios, and homepages via Exa's personal-site index — a person's own voice rather than coverage of them. Calls the Exa REST API through a local script; no MCP server needed, only an EXA_API_KEY.

Setup (once)

bash
export EXA_API_KEY=your-key            # or EXA_API_KEY=... in a .env at the repo root

Get a key at https://dashboard.exa.ai/api-keys. Shared details: exa-native-base.

Run it

bash
python scripts/personal_site_search.py "<who / what they write about>" [-n N] [--text]

Examples:

bash
python scripts/personal_site_search.py "machine learning engineer personal blog" -n 10
python scripts/personal_site_search.py "indie hacker building in public portfolio" -n 12
python scripts/personal_site_search.py "Andrej Karpathy personal site essays" -n 5

Defaults to category=personal site.

Query patterns

bash
# Practitioners who write about a topic
python scripts/personal_site_search.py "category:personal site distributed systems engineer blog" -n 12

# A specific person's own site
python scripts/personal_site_search.py "category:personal site <name> blog essays projects" -n 5

# Portfolios in a discipline
python scripts/personal_site_search.py "category:personal site product designer portfolio case studies" -n 12

Hidden relationships & terminology

Personal sites are ideal for finding non-obvious connections (collaborators, longtime clients, mentors) that aren't listed anywhere. Use indirect signals:

bash
python scripts/personal_site_search.py "<subject> conversation interview testimonial guest" -n 8
python ../exa-native-base/scripts/exa.py search "<subject> years decades longtime worked with known since" -n 10

Detect insider terminology on the subject's own pages, then search for others using those terms. (See the relationships pattern in exa-lead-generation.)

Override categories with -c

personal site (default) · linkedin profile (their professional profile) · news (press about them).

After you get results

• This index favours genuine personal sites; still treat results as similarity, not validation and open the page to confirm it's the right person.
• Deep-read with python ../exa-native-base/scripts/exa.py contents <url> --text to assess voice/expertise.
• Deliver: name · site URL · what they write about · a representative post.