MCP App Store
by KeenEthics
The Swarm icon

The Swarm MCP Server for ChatGPT

By The Swarm

Overview

The Swarm gives AI assistants direct access to professional profile and company intelligence plus relationship-network mapping. Retrieve employment history, education, skills, locations, company details, funding information, and career progression. Search a team's professional network to identify connection owners and relationship strength, or map a partner organization's network when the account has partner access. Built for sales prospecting, recruiting, partnership development, and competitive intelligence. Access is limited to data the signed-in Swarm account is authorized to view.

How to connect

  1. ChatGPT: Open the The Swarm listing in the ChatGPT app directory and choose Connect. ChatGPT listing The Swarm exposes 9 tools here.

Tools

company_dictionary

ChatGPT
Look up the exact stored values of a filterable field (dictionary) in the companies index - e.g. how an industry or location is spelled - before using them in "term" filters. Refer to the companies index mapping to see if the field is mapped as "text" or "keyword": - "text" fields: terms are prefix-matched (<term>) and can be combined with "AND", "OR", and "NOT" in a single call. - "keyword" fields: the query is wildcard-matched (<term>*); look up one term per call. The scope dictionary always returns its two fixed values (inNetwork, outOfNetwork) regardless of the query. Example: field="industry", query="cyber" returns values like "Cybersecurity" - use these exact strings in "term" filters.

find_companies

ChatGPT
Search The Swarm's company database with an OpenSearch query. Returns full company records: industries, size and workforce, locations, funding rounds, business status, and social links. Workflow: 1. Call get_company_mapping once per conversation before first use. 2. Resolve exact filter values with company_dictionary (industry, location, etc.). 3. Build the query and call this tool. Query rules: - Fields mapped as "nested" (e.g. company_info.locations, funding_data.rounds) require a "nested" query with the full path. - Use "term"/"terms" with exact dictionary values on "keyword" fields; use "match" on "text" fields. - Querying a field that is not in the mapping matches nothing and returns an EMPTY result, not an error. When this is detected, the response contains a "warnings" array naming the unmapped fields. If you get zero results unexpectedly, check "warnings" and re-check field names against the mapping. Example - US cybersecurity companies (keyword term + nested location): {"bool": {"filter": [{"term": {"company_info.industry": "cybersecurity"}}, {"nested": {"path": "company_info.locations", "query": {"term": {"company_info.locations.country": "united states"}}}}]}} Results are sorted by company size (largest first), then relevance. Each company record is very large (5-25KB of JSON). ALWAYS set "limit" explicitly and keep it small - 10 or fewer unless the task genuinely needs more. The "query" argument uses the OpenSearch query DSL accepted by The Swarm's company search API, documented at https://docs.theswarm.com/docs/endpoints/v3/search-companies.

find_profiles

ChatGPT
Search The Swarm's people database with an OpenSearch query. Returns full profiles: current and past employment, education, skills, locations, emails, and social links. Workflow: 1. Call get_profile_mapping once per conversation before first use. 2. Resolve exact filter values with profile_dictionary (industries, locations, job titles, etc.). 3. Build the query and call this tool. Query rules: - Fields mapped as "nested" (e.g. profile_info.experience, connections) require a "nested" query with the full path. Non-nested profile_info.* fields are queried directly. - Use "term"/"terms" with exact dictionary values on "keyword" fields; use "match" on "text" fields (job titles, company names, locations). - Querying a field that is not in the mapping matches nothing and returns an EMPTY result, not an error. When this is detected, the response contains a "warnings" array naming the unmapped fields. If you get zero results unexpectedly, check "warnings" and re-check field names against the mapping before concluding no one matches. Examples: - Founders (text match): {"match": {"profile_info.job_title": "founder"}} - US-based executives (keyword terms, values from profile_dictionary): {"bool": {"filter": [{"term": {"profile_info.job_seniority": "cxo"}}, {"term": {"profile_info.job_company_location_country": "united states"}}]}} - People who have ever worked at a company (nested field): {"nested": {"path": "profile_info.experience", "query": {"term": {"profile_info.experience.company.website": "anthropic.com"}}}} Each returned profile is large (10-15KB of JSON). Keep "limit" as low as the task allows; start small when exploring. The "query" argument uses the OpenSearch query DSL accepted by The Swarm's profile search API, documented at https://docs.theswarm.com/docs/endpoints/v3/search-profiles.

get_company_mapping

ChatGPT
Get the OpenSearch mapping (fields, types, analyzers) of the companies index. Call this once before the first use of find_companies, then reuse it - the mapping does not change within a conversation. Reading the mapping: - Only fields present in the mapping are searchable (dynamic mapping is off); any other field silently matches nothing. - Fields of type "nested" (e.g. company_info.locations, funding_data.rounds) must be queried with a "nested" query. - "keyword" fields with a lowercase normalizer accept any casing in term queries.

get_profile_mapping

ChatGPT
Get the OpenSearch mapping (fields, types, analyzers) of the profiles index. Call this once before the first use of find_profiles or relationships or partner_network_mapper, then reuse it - the mapping does not change within a conversation. Reading the mapping: - Only fields present in the mapping are searchable (dynamic mapping is off); any other field silently matches nothing. - "alias" fields (e.g. profile_info.current_title) point at the field in their "path"; either name works in queries. - Fields of type "nested" (e.g. profile_info.experience, connections) must be queried with a "nested" query. - "keyword" fields with a lowercase normalizer accept any casing in term queries.

network_mapping_status

ChatGPT
Requires The Swarm partner access - if this tool returns "forbidden", tell the user their account lacks partner permissions and stop; do not retry. Get the current status of an asynchronous network mapping task started by the "partner_network_mapper" tool. Once the status indicates mapping is complete, call "partner_network_mapper" again with the original arguments to retrieve the results.

partner_network_mapper

ChatGPT
Requires The Swarm partner access - if this tool returns "forbidden", tell the user their account lacks partner permissions and stop; do not retry. Search people in ANOTHER company's network by OpenSearch profile query. The target company is identified by its website domain. Same query format and rules as the "relationships" tool: call get_profile_mapping first; unmapped fields silently match nothing. - If the target network is already mapped: returns results in the same shape as the "relationships" tool (max 50 items plus "totalCount"). - If not yet mapped: starts an asynchronous mapping task and returns "taskId" only. Check progress with "network_mapping_status"; mapping can take a while, so inform the user and avoid tight polling. Once mapping is finished, call this tool again with the same arguments to get the results. The "query" argument uses the OpenSearch query DSL accepted by The Swarm's network mapper API, documented at https://docs.theswarm.com/docs/endpoints/v3/network-mapper.

profile_dictionary

ChatGPT
Look up the exact stored values of a filterable field (dictionary) in the profiles index - e.g. how an industry, location, or job title is spelled - before using them in "term" filters. Refer to the profiles index mapping to see if the field is mapped as "text" or "keyword": - "text" fields: terms are prefix-matched (<term>) and can be combined with "AND", "OR", and "NOT" in a single call. - "keyword" fields: the query is wildcard-matched (<term>*); look up one term per call. The seniority dictionary is small; known values include: cxo, c-level, vp, director, manager, senior, entry, owner, training. You can use these directly, but verify with this tool if a term filter unexpectedly returns nothing. Fixed-value dictionaries - the full value list is always returned and the query text is ignored (any valid query works): - connectionStrength, teamConnectionStrength: normalized strength "0" (unfamiliar) to "3" (strong/confirmed), used on connections.strength_normalized. - connectionSource: how a relationship was established (LinkedIn import, Google contacts/calendar, work overlaps, ...). - pipelineStatus: pipeline stage values (suggested, introRequested, inContact, meetingSet, removed). Example: field="industry", query="software" returns values like "Software & Saas", "Computer Software", "Enterprise Software" - use these exact strings in "term" filters.

relationships

ChatGPT
Search people in YOUR team's network (people your team members have a relationship with) using an OpenSearch profile query. Same query rules as find_profiles: call get_profile_mapping first; unmapped fields silently match nothing. Returns up to "limit" items plus "totalCount" of all matches. If the response includes a "pagination_token", pass it in the next call with the same query to retrieve the next page. Each item contains: - profile: name, current title, company, LinkedIn URL, work email. - connections: the team members who know this person, each with connection_strength (0-1 float), connection_strength_normalized (1=weak to 3=strong), and sources[].origin (how they know each other, e.g. linkedin_connection, calendar_events). Use this tool for "who do we know at/that..." questions. Use find_profiles with inNetworkOnly=true instead when you need full profile details. Example - who does my team know at Anthropic: {"match": {"profile_info.job_company_name": "anthropic"}} The "query" argument uses the OpenSearch query DSL accepted by The Swarm's relationships API, documented at https://docs.theswarm.com/docs/endpoints/v3/relationships.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/the-swarm"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/the-swarm)