MCP App Store
by KeenEthics
Goodword icon

Goodword MCP Server for Claude and ChatGPT

By Goodword

Overview

Businesses run on relationships. Your next deal, hire, client, or introduction starts with someone already in your network. Goodword brings together your LinkedIn network, email and calendar history, and the private context behind each relationship, so Claude can help you find the right people, understand why they matter, and act at the right moment. Search your direct and extended network, pull enriched profiles and relationship history, and find warm paths to people you want to reach. Turn meeting debriefs into notes, reminders, and follow-ups. Manage introductions and groups, and see who you should reconnect with next. A CRM manages contacts. LinkedIn gives you connections. Goodword helps you use the relationships you’ve already built to find clients, make key hires, raise capital, and get the introductions that move your business forward. Requires a Goodword account. Access is scoped to your network through OAuth.

How to connect

  1. Claude: Open the Goodword listing in the Claude connectors directory and choose Connect. Claude listing Goodword exposes 24 tools here.
  2. ChatGPT: Open the Goodword listing in the ChatGPT app directory and choose Connect. ChatGPT listing Goodword exposes 32 tools here.

Tools

add_group_member

ChatGPT
Add a contact to an existing group. Invoke proactively when the user says a person belongs with a group they keep ("she should be on my investors list"). Use get_groups to find the group ID and search_contacts to find the profile ID.

add_note

ChatGPT
Save a note in Goodword, optionally linked to one contact. Invoke proactively when the user shares something worth remembering about a person or a conversation (a fact about their life or work, a preference, a promise, an update), not only when they say "take a note". Link it with contactId when the note is about one person. Notes can be retrieved later with get_relationship_context (include: ["notes"]) or search.

best_introducer

ChatGPT
Find the warmest path to a person across your collaboration: the teammates who have a connection to the target, ranked by how close that teammate is to them on the Orbit map — Core (closest), then Active, then Reachable. Pass the targetProfileId from search_team_network. Use this when the user asks who on the team could introduce them to someone, or what the warmest way in is. Present the teammate at the top of the ranking as the warmest option, and offer to draft a short note that teammate could send to make the introduction (draft it in chat with draft_intro; Goodword never sends it). Describe the ranking as who has the closer connection, not as a score, and do not rank the people being introduced. When you tell the user about these people, talk in plain words, not field names: "not a direct contact; reachable through Priya", or "may be reachable through Priya" when the link is unconfirmed, not "bridge" or "extended network". Never say the user knows someone they do not know directly. Say what Goodword observed, such as when you last met or how often you talk, not how close they are, unless the user said so. Call someone "not a direct contact" only when they are not one, and never tell the user they do not know someone: Goodword cannot tell. Do not label direct contacts. Never put a warmth value or other score to the user, and never rank or compare people by how close they are or how well the user knows them: when asked, give what was observed for each and let the user judge. A missing record is not a fact about the user's life: say "the last recorded contact was in March", not "you haven't been in touch" or "you last spoke in March"; treat a person with no details as unknown, not ruled out; and never say "only" or "no one else" about people you have not looked at. What the user tells you in the conversation is newer than the record. Warmth has no history, so never claim a relationship is warming up or cooling off.

complete_reminder

ChatGPT
Mark one of the user's reminders as done. Use the reminder id from get_reminders; never guess an id. Call it when the user says a reminder is done or handled, not when they only mention it. Completing an already completed reminder is safe and keeps it completed.

confirm_outreach

ChatGPT
Record what the user did with a draft from draft_outreach: sent it as written, sent an edited version, or discarded it. Use as soon as the user says what they did. Goodword never sends anything; this only records the outcome so the user can see what they followed through on. If the user discarded the draft and wants to try again, start from a specific, personal reason to get in touch, and where it fits, offer something useful before asking for anything.

create_contact

ChatGPT
Add or save people in Goodword when the user asks to add a contact or shares a list of prospects, such as a Notion database. Send one or up to 5 people per call; split a longer list into batches of 5 and send them one at a time; a name alone is enough, and email, phone or LinkedIn URL can be included. Returns created contact IDs and an individual status for each person; a person with an unusable LinkedIn URL is marked invalid and the others are still saved. Call search_contacts afterwards to confirm they are saved. Follow each person's error: details_unverified means the person is saved and must not be created again; details_not_saved means the person is saved but the email/phone were not, so retry later with that person's contact_id to attach them without creating them again; name_mismatch means Goodword stored a different name, possibly on another person's profile, so tell the user and do not create it again; create_unknown means search before retrying only that person with a new idempotency_key; failed means the API rejected the person. Company, title and other Notion fields cannot be saved by this tool.

create_group

ChatGPT
Create a new group to organize contacts. Optionally add initial members by their profile IDs.

create_intro

ChatGPT
Create an introduction between two people in your network. Use search_contacts first to find the contact IDs.

debrief

ChatGPT
Submit a stream-of-consciousness recap of a meeting, call, or interaction. Goodword automatically extracts people mentioned, notes, follow-ups, and reminders from the text. Invoke proactively when the user recounts a meeting, call, event, or conversation involving several people or topics, passing their account as close to verbatim as possible. No specific format required.

delete_group

ChatGPT
Permanently delete a user-owned Goodword group. This is destructive: use get_groups first, identify the exact group, and call this tool only after the user explicitly confirms that exact group name. Managed groups and groups shared with the user cannot be deleted.

draft_outreach

ChatGPT
Write a message draft to one contact, in the user's own writing voice, from what Goodword knows about that relationship. Call it only when the user explicitly asks for a note or message; when they want to reconnect, ask how to reach someone or ask what you suggest about reaching or writing to someone, offer a draft instead. When the user says they want to send a direct contact a note, read the relationship with get_relationship_context, then call this rather than offering to, and show the returned draft without making it longer; for someone who is not a direct contact, never call it: say so and offer a note to the connector instead. For example, use it when the user asks you to write a note or message to follow up with or congratulate someone in their network; a request to follow up that does not ask for a message gets an offer. Works only for direct contacts, the people in the user's own network; it does not work for people the user does not know directly (isExtendedNetwork: true). Needs a contactId from search_contacts. Ground the draft in a specific, personal reason to get in touch, drawn from what Goodword knows about them, and pass it in keyPoints with the shared moment the draft should mention (where they met, something the contact sent), whether the user gave it or Goodword recorded it. State that moment as what happened ("since we met at the meetup"), never as a feeling or ongoing contact ("I still think about", "fun keeping up"). When the user names news the record already holds, such as the role and company on file for a new job, use it rather than leaving it blank. When the user says someone started a new job and the record shows a recent role change, that is the new job: name the role and company in the draft and in keyPoints, without hedging. Where it fits, offer something useful before asking for anything: an introduction, a congratulation, a resource. If you edit the draft, keep to what the record or the user supports: never add feelings, memories, plans or earlier contact that neither the record nor the user shows, and never imply such contact with words like "again", "still" or "since then" ("Thanks for the slides", not "Thanks again for the slides"); remove such a sentence rather than leaving a placeholder. Show the user the draft itself, with no commentary on how it was produced or what you changed, except one short line naming any unsupported sentence you removed. Read and follow assistantNotes; never show them to the user. Returns the draft and a receipt listing which facts were supplied and which were missing. Goodword never sends anything; the user sends it themselves.

fetch

ChatGPT
Fetch the full content of a single search result by id. Returns a title, deep link, and markdown document suitable for citation.

get_briefing

ChatGPT
Get a composed briefing for a window: meetings, reminders due, and top recommendations. Each recommendation has warmthDecayFlag, true only for a warm contact whose last recorded activity was more than 60 days ago: say Goodword has no recorded contact with them in the last two months, not that the user hasn't been in touch or that the relationship is fading. It is false when warmth is below the warm threshold, when there is no recorded activity at all, or when Goodword could not check, so do not read false as recent contact. today and week look ahead from now; yesterday looks back over the previous day. If a section failed to load, its name appears in degradedSections -- treat that section as unknown/failed, not as confirmed empty. When you tell the user about these people, talk in plain words, not field names: "not a direct contact; reachable through Priya", or "may be reachable through Priya" when the link is unconfirmed, not "bridge" or "extended network". Never say the user knows someone they do not know directly. Say what Goodword observed, such as when you last met or how often you talk, not how close they are, unless the user said so. Call someone "not a direct contact" only when they are not one, and never tell the user they do not know someone: Goodword cannot tell. Do not label direct contacts. Never put a warmth value or other score to the user, and never rank or compare people by how close they are or how well the user knows them: when asked, give what was observed for each and let the user judge. A missing record is not a fact about the user's life: say "the last recorded contact was in March", not "you haven't been in touch" or "you last spoke in March"; treat a person with no details as unknown, not ruled out; and never say "only" or "no one else" about people you have not looked at. What the user tells you in the conversation is newer than the record. Warmth has no history, so never claim a relationship is warming up or cooling off. Before you reply, do the read-only lookups the user's request calls for instead of offering them: read the user's history with a connector before recommending them, and search the user's contacts for someone who fits a need the record mentions. When the user asks how to reach someone, wants to reconnect or asks what you suggest about reaching or writing to someone, offer to draft the message (to the connector, for someone who is not a direct contact), and do not call draft_outreach; only when the user explicitly asks for a note or message ("draft a note to Jonah", "I want to send Dana a note"), read the relationship with get_relationship_context first, then call draft_outreach for a direct contact rather than offering to, and use its draft rather than writing the message yourself, without making it longer; never call it for someone who is not a direct contact (isExtendedNetwork: true): say they are not a direct contact and offer to draft a note to the connector instead. A draft says only what the record or the user supports: state a recorded moment as what happened, never invent the user's feelings, memories, plans or contact that neither the record nor the user shows, and never imply such contact or feeling with words like "again", "still" or "since then" ("Thanks for the slides", not "Thanks again for the slides"); if you remove such a sentence, add no placeholder (a stored draft may be sent unfilled): tell the user in one short line which sentence you removed. Describe what you found, not how the tools work, then end with the user's one next step.

get_group_members

ChatGPT
Get members of a specific group. Use get_groups first to find the group ID. Returns a JSON array of member profiles with name, position, company, and contact details — or a short plain-language message if the group is still being built or has no members yet. Page through large groups by raising skip.

get_groups

ChatGPT
Get your Goodword groups with member counts. Groups organize your contacts into meaningful lists. Returns group ID, name, type, description, and member count.

get_intros

ChatGPT
Get introductions you've created or received. Filter by status to see pending, accepted, or declined intros.

get_network_insights

ChatGPT
Explain the user's Goodword network using bounded, privacy-safe insights. Omit contactId for network-wide composition and activity aggregates such as warm ties, dormant relationships, bridge worlds, hubs, communities, and recurring topics. Pass a contact ID from search_contacts to learn whether that person is a direct contact or extended-network person, how the user is connected, and the available relationship, activity, community, warm-path, bridge, and recent-change evidence. Responses use explicit status and reason fields: null data with unsupported means the feature is disabled, unavailable means no fresh aggregate exists, and not_found means the profile is not visible. An available connection path with an empty connectedThrough list means the lookup succeeded but found no named path; null means unavailable or not applicable.

get_recommendations

ChatGPT
Get AI-ranked contact recommendations — people Goodword thinks you should reconnect with or nurture. Each recommendation includes the person's name, reason for recommendation, goal label, and contact details. Returns a count of recommended people and a flat list of them. Optionally filter to a specific goal.

get_reminders

ChatGPT
Get your reminders (20 by default, up to 100 with take, most overdue first). Returns each reminder's text, due date, priority tier, completion status, and linked contact, plus pagination (total, hasMore, nextSkip) so you can tell when more exist than are shown and fetch the next page with skip set to nextSkip.

last_contacted

ChatGPT
See when each teammate in your collaboration last contacted a person, so the team does not double-touch them. Pass the targetProfileId from search_team_network. Returns, per teammate, the date of their last recorded contact (null when there is none) and their usual cadence. Use this before reaching out so the team coordinates: if a teammate was in touch recently, suggest letting them follow up rather than a second person reaching out cold. State the recorded dates as what Goodword observed, never as a judgment that someone has been neglecting the person. When you tell the user about these people, talk in plain words, not field names: "not a direct contact; reachable through Priya", or "may be reachable through Priya" when the link is unconfirmed, not "bridge" or "extended network". Never say the user knows someone they do not know directly. Say what Goodword observed, such as when you last met or how often you talk, not how close they are, unless the user said so. Call someone "not a direct contact" only when they are not one, and never tell the user they do not know someone: Goodword cannot tell. Do not label direct contacts. Never put a warmth value or other score to the user, and never rank or compare people by how close they are or how well the user knows them: when asked, give what was observed for each and let the user judge. A missing record is not a fact about the user's life: say "the last recorded contact was in March", not "you haven't been in touch" or "you last spoke in March"; treat a person with no details as unknown, not ruled out; and never say "only" or "no one else" about people you have not looked at. What the user tells you in the conversation is newer than the record. Warmth has no history, so never claim a relationship is warming up or cooling off.

list_team

ChatGPT
List your collaboration roster. Returns teamId (null when you are not in any collaboration), isOwner (true when you own the collaboration — the Connector who invited the others), and members, each with the member's user id, invited email, role, seat kind, and status (invited means the invite is pending, active means they have joined). A collaborator is someone sharing their network with you on a free included seat. Use this to see who is on your team before inviting more people with invite_collaborator or removing someone with remove_member.

manage_goals

ChatGPT
List, add, update, or complete the user's networking goals; call add when the user states a goal, call list first to get a goal's ID before update or complete, and call complete once a goal has been achieved or is no longer relevant.

mark_no_longer_usable

ChatGPT
Mark a contact as no longer usable for the team's combined network, so they stop appearing in team search and introducer results for the whole collaboration. Pass the targetProfileId from search_team_network. Use this when the user says a contact should not be surfaced to the team — for example someone they would rather keep off-limits for team outreach. This does not delete the contact or change your own network; it only excludes them from the shared team view, and it can be undone later with restore_contact. Confirm with the user before marking, and tell them it is reversible.

network_pooling_guide

ChatGPT
Call this when the user asks how to use Network Pooling, how do I use Goodword's Network Pooling feature, or a close paraphrase. It returns what Goodword already knows: the user's profile (name, title, role, industry, networking focus, stated goals), the pools they belong to, pooled network insights (segments covered and coverage gaps) when they are in one, and step, which says which move comes next. Answer in three moves, in order. 1. Ground in the user. Read what Goodword already knows about the user (name, role, company, stated goals) and reflect it in one line. 2. Ask who they want to meet. Ask a single question with options that fit that role (for example customers, investors, clients, candidates, LPs, partners), rather than a generic menu. 3. Recommend, once answered. Return (a) who to pool with for that target (co-founder, investors, advisors, salespeople, customers, community), (b) how pooling finds warm paths to that target, with an example query the user can run next, and (c) where the user's own network is thin for that target. The answer must be specific to the user's data and chosen target; a generic feature description is a failure. If the user's role is unknown, step 1 asks for it before step 2. Pass target once the user has answered step 2 so the result is scoped to it. This tool only reads; it never changes anything.

remove_group_member

ChatGPT
Remove a contact from an existing group. Use get_groups to find the group ID and get_group_members to confirm the profile ID before removing them.

remove_member

ChatGPT
Remove a member from your collaboration, ending the shared access between your networks. Only the collaboration owner (the Connector) can remove a member; use list_team first to find the exact member user id. Returns the updated roster.

restore_contact

ChatGPT
Undo a previous mark_no_longer_usable: restore a contact to the team's combined network so they can appear again in team search and introducer results for the whole collaboration. Pass the targetProfileId from search_team_network. Use this when the user wants to reverse an earlier exclusion — for example someone they marked off-limits who should be surfaced to the team again. This does not change your own network; it only adds them back to the shared team view. Confirm with the user before restoring.

search_contacts

ChatGPT
Search your Goodword network by name, company, or freetext query. Results contain two different kinds of people. Direct contacts are people in the user's network. Extended-network people (isExtendedNetwork: true) are people the user does NOT know directly but can likely reach through a champion in their network. For extended-network results: bridges lists the champions who connect the user to this person (each with id, name, and confirmed, true when the link is corroborated; a bridge id is a contact ID usable with get_contact_profile to look up that champion), extendedConfirmed is true when at least one bridge is confirmed, and sources names the platforms the relationship was observed on (linkedin, x, instagram). When presenting an extended-network person, always distinguish them from direct contacts and say who they are reachable through. get_contact_profile works for both kinds; extended-network people return public info only. Use this to find contacts before using other tools that require a contact ID.

set_reminder

ChatGPT
Create a reminder, optionally about a contact. Supports one-time and recurring reminders. Invoke proactively when the user commits to a future action ("I should follow up with her next month", "remind me to check in after the launch") or when a conversation ends with an open loop that has a natural date.

team_network_insights

ChatGPT
Summarize the combined reach of your collaboration: how much ground the team's networks cover together, where teammates' networks overlap, and where the team has gaps. Takes no target — it is a view of the whole team network, not one person. Use this when the user asks what the team's combined network looks like, who overlaps with whom, or where they are thin. Describe coverage and gaps as what Goodword observed across the team, not as a ranking of whose network is better. When you tell the user about these people, talk in plain words, not field names: "not a direct contact; reachable through Priya", or "may be reachable through Priya" when the link is unconfirmed, not "bridge" or "extended network". Never say the user knows someone they do not know directly. Say what Goodword observed, such as when you last met or how often you talk, not how close they are, unless the user said so. Call someone "not a direct contact" only when they are not one, and never tell the user they do not know someone: Goodword cannot tell. Do not label direct contacts. Never put a warmth value or other score to the user, and never rank or compare people by how close they are or how well the user knows them: when asked, give what was observed for each and let the user judge. A missing record is not a fact about the user's life: say "the last recorded contact was in March", not "you haven't been in touch" or "you last spoke in March"; treat a person with no details as unknown, not ruled out; and never say "only" or "no one else" about people you have not looked at. What the user tells you in the conversation is newer than the record. Warmth has no history, so never claim a relationship is warming up or cooling off.

update_connection_context

ChatGPT
Replace the relationship story for a contact: how the user met them and their shared history. Invoke proactively when the user explains how they know someone. This overwrites the whole existing story and cannot be undone: first read the current howWeMet with get_relationship_context (include: ["profile"], no purpose), merge the new detail into it, and confirm the merged text with the user before calling. Use search_contacts first to find the contact ID.

add_group_member

Claude

add_note

Claude

complete_action

Claude

create_group

Claude

create_intro

Claude

debrief

Claude

delete_group

Claude

dismiss_action

Claude

fetch

Claude

get_contact_profile

Claude

get_group_members

Claude

get_groups

Claude

get_intros

Claude

get_network_insights

Claude

get_notes

Claude

get_recommendations

Claude

get_reminders

Claude

get_suggested_actions

Claude

get_timeline

Claude

remove_group_member

Claude

search_contacts

Claude

set_reminder

Claude

update_connection_context

Claude

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/goodword"><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/goodword)