List Building

Build company lists using Extruct API methods, guided by a decision tree. Reads from the company context file for ICP and seed companies.

Official API Reference

• https://www.extruct.ai/docs

Decision Tree

Before running any queries, determine the right approach:

Have a seed company from win cases or context file?
  YES → Method 1: Lookalike Search (pass seed domain)
  NO  ↓

New vertical, need broad exploration?
  YES → Method 2: Semantic Search (3-5 queries from different angles)
  NO  ↓

Need qualification against specific criteria?
  YES → Method 3: Discovery API (criteria-scored async research)
  NO  ↓

Need maximum coverage?
  YES → Combine Search + Discovery (~15% overlap expected)

Before You Start

Read the company context file if it exists:

claude-code-gtm/context/{company}_context.md

Extract:

• ICP profiles — for query design and filters
• Win cases — for seed companies in lookalike mode
• DNC list — domains to exclude from results. If no DNC list exists in the context file, ask the user: (1) run an Extruct search for competitors to auto-populate, (2) accept a CSV of existing customers/partners, or (3) skip for now

Also check for a hypothesis set at claude-code-gtm/context/{vertical-slug}/hypothesis_set.md. If it exists, use the Search angle field from each hypothesis to design search queries — these are pre-defined query suggestions tailored to each pain point.

Environment

Variable	Service
EXTRUCT_API_TOKEN	Extruct API

Before making API calls, check that EXTRUCT_API_TOKEN is set by running test -n "$EXTRUCT_API_TOKEN" && echo "set" || echo "missing". If missing, ask the user to provide their Extruct API token and set it via export EXTRUCT_API_TOKEN=<value>. Do not proceed until confirmed.

Base URL: https://api.extruct.ai/v1

Method 1: Lookalike Search

Use when you have a seed company (from win cases, existing customers, or user input).

Endpoint: GET /companies/{identifier}/similar where identifier is a domain or company UUID.

Key params:

• filters — JSON with include (size, country) and range (founded)
• limit — max results (default 100, up to 200)
• offset — for pagination

Response fields: name, domain, short_description, founding_year, employee_count, hq_country, hq_city, relevance_score

When to use lookalike:

• You have a happy customer and want more like them
• Context file has win cases with domains
• User says "find companies similar to X"

Tips:

• Run multiple similar-company searches with different seed companies for broader coverage
• Combine with filters to constrain geography or size
• Deduplicate across runs by domain
• Default to limit=100; increase up to 200 when broader coverage is needed

Method 2: Semantic Search — Fast, Broad

Endpoint: GET /companies/search

Key params:

• q — natural language query describing the target companies
• filters — JSON with include (size, country) and range (founded)
• limit — max results (default 100, up to 200)

Response fields: name, domain, short_description, founding_year, employee_count, hq_country, hq_city, relevance_score

Query strategy:

• Write 3-5 queries per campaign, each from a different angle on the same ICP
• Describe the product/use case, not the company type
• Deduplicate across queries by domain — overlap is expected
• Default to limit=100 per query; increase up to 200 when needed
• Target 200-800 companies total across all queries

Filters: See references/search-filters.md

Method 3: Discovery API — Deep, Qualified

Endpoint: POST /discovery_tasks

Key params:

• query — 2-3 sentence description of the ideal company (like a job description)
• desired_num_results — target result count (default 50 for first pass)
• criteria — list of { key, name, criterion } objects for auto-grading (up to 5)

Poll: GET /discovery_tasks/{task_id} — status: created | in_progress | done | failed. Poll every 60 seconds.

Fetch results: GET /discovery_tasks/{task_id}/results with limit and offset params.

Response fields: company_name, company_website, company_description, relevance (0-100), scores (per-criteria grade 1-5 with explanation), founding_year

Query strategy:

• Write queries like a job description — 2-3 sentences describing the ideal company
• Use criteria to auto-qualify — each company gets graded 1-5 per criterion
• Default desired_num_results=50 for first pass; expand after quality review
• Use up to 5 criteria per task; keep criteria focused and non-overlapping
• Run separate tasks for different ICP segments
• Scans many candidates to find qualified matches — runtime depends on query scope
• Up to 250 results per task

See references/discovery-api.md for full parameters.

Upload to Table

Create a company kind table via POST /tables with a single input column (kind: "input", key: "input"). Extruct auto-enriches each domain with a Company Profile.

Upload domains in batches of 50 via POST /tables/{table_id}/rows. Each row: { "data": { "input": "domain.com" } }. Add 0.5s delay between batches.

Pass "run": true in the rows payload to trigger agent columns on upload.

Re-run After Enrichment

After the list-enrichment skill adds data points to this list, consider re-running list building using enrichment insights as Discovery criteria. For example:

• If enrichment reveals that "companies using legacy ERP" are the best fit, create a Discovery task with that as a criterion
• If enrichment shows a geographic cluster, run a Search with tighter geo filters
• This creates a feedback loop: list → enrich → learn → refine list

Result Size Guidance

Campaign stage	Target list size	Method
Exploration	50-100	Search (2-3 queries)
First campaign	200-500	Search (5 queries) + Discovery
Scaling	500-2000	Discovery (high desired_num_results) + multiple Search

Workflow

Step 0: Verify API reference

1. Read local references: references/discovery-api.md, references/search-filters.md
2. Fetch live docs: https://www.extruct.ai/docs
3. Compare endpoints, params, and response fields
4. If discrepancies found:
   • Update the local reference file(s)
   • Flag changes to the user before proceeding
5. Proceed with the skill workflow

Steps

1. Read context file for ICP, seed companies, and DNC list
2. Follow the decision tree to pick the right method
3. Draft queries (3-5 for Search, 1-2 for Discovery)
4. Run queries and collect results
5. Deduplicate across all results by domain
6. Remove DNC domains
7. Upload to Extruct company table for auto-enrichment
8. Add agent columns if user needs custom research
9. Ask user for preferred output: Extruct table link, local CSV, or both