MCP App Store
by KeenEthics
Gojiberry icon

Gojiberry MCP Server for ChatGPT

By SUPERFRUITS

Overview

Gojiberry brings your outbound sales workflow into ChatGPT. Search a directory of B2B leads and companies with free masked previews, reveal full records, and enrich contacts with verified emails and mobile phone numbers. Organize contacts into lists, create and manage outreach campaigns, review campaign logs, configure lead-sourcing Agents, and read and reply to conversations in your unified inbox - all through natural language. Requires a Gojiberry account and API key from app.gojiberry.ai.

How to connect

  1. ChatGPT: Open the Gojiberry listing in the ChatGPT app directory and choose Connect. ChatGPT listing Gojiberry exposes 51 tools here.

Tools

add_contact_signal

ChatGPT
Add a signal to a contact, shown next to the signals detected by Gojiberry sources (most recent first). Designed for first-party signals: something the lead did in the user's own product or stack (signed up for a trial, booked a demo, attended a webinar, downloaded the pricing guide, ...). First-party signal: describe what happened in intent (e.g. 'Booked a demo from the pricing page'), optionally add a short label in intentKeyword (e.g. 'demo_booked'), and leave intentType empty. Signals without intentType are recorded on every call, so calling twice records the event twice — call it once per event. Set intentType only to record one of the signal types Gojiberry sources detect. A typed signal the contact already has (same intentType and intentKeyword) is not recorded again and the response has created: false; HIRING, HIRING_SURGE, NEW_DECISION_MAKER, RECENT_ACTIVITY, RECENTLY_CHANGED_JOB and RECENT_FUNDING_EVENT are recorded at most once per contact, whatever the keyword. At least one of intent, intentType or intentKeyword is required. Returns created and the contact with its up-to-date signals.

add_contacts_to_list

ChatGPT
Add existing contacts to a list. In Gojiberry, campaign membership is driven by list membership: adding a contact to a list that is linked to a campaign enrolls the contact in that campaign.

create_agent

ChatGPT
Create a new source agent — a lead finder that continuously sources leads matching an ICP, scores them with AI, and imports them into a list. Source agents do NOT do outreach; outreach is done by a campaign agent (create_campaign). IMPORTANT — a list is mandatory: the agent imports the leads it sources into a list, so you must create a list first (use create_list) and pass its ID in the required listId field. The list must belong to the authenticated user. FULL CYCLE — a source agent alone is almost never what the user wants: without a campaign it shows up as a 'legacy agent' in the platform, which only sources leads and does no outreach. Guide the user to the Full Cycle setup (leads sourced AND outreached automatically, end to end) by creating three resources in this order: 1. create_list — the list is the link between the two agents; 2. create_agent (this tool) — the source agent, with that list as listId; 3. create_campaign — the campaign agent, with that list in listIds. For agentType: "autopilot" (the default), signal variables configure where leads are sourced from. When variables are provided, at least 4 are required (max 15). Each variable must have a value, which usually contains a keyword or a piece of data matching its type (e.g. a search keyword for SEARCH_KEYWORD, a LinkedIn page URL for COMPETITOR_PAGE_URL). LOOKALIKE variables are not allowed, and JOB_SEARCH variables are not authorized for now. If the active agent limit is reached, the agent is still created but paused.

create_campaign

ChatGPT
Create an outreach campaign (the campaign agent — what the Gojiberry platform calls an 'Agent') with its steps, lead-source lists, and sending seats. The campaign reaches out to the contacts of its lists. FULL CYCLE — this is the last of the three resources of a Full Cycle agent (leads sourced AND outreached automatically, end to end), created in this order: 1. create_list — the list linking the two agents; 2. create_agent — the source agent that fills the list with leads; 3. create_campaign (this tool) — the campaign agent, with that list in listIds. Without this campaign, a source agent is just a 'legacy agent' that sources leads but does no outreach. SENSITIVE — unless the user explicitly asked to launch the campaign immediately, pass active: false. The API defaults active to true, and an active campaign starts the outreach automatically on a daily basis (launch hours are configured in the LinkedIn seat settings). Lists (lead source): - At least one list is required (listIds). The campaign reaches out to the contacts of its lists. Lists are typically filled by a source agent (see create_agent) or by importing contacts (add_contacts_to_list). Steps — executed in array order; stepNumber is assigned automatically from the order. Valid types: invitation, invitationNote, message, voiceMessage, visitProfile, likePosts (all LinkedIn), and email. Ordering rules (the API rejects violations): - At most ONE Connection Request step (invitation or invitationNote) per campaign. - An invitation cannot come AFTER a message or voiceMessage step. - Every step's delayAfterLastStep must be at least 1 day (integer, default 2). Per-type fields: - message and email steps take a messageMode: 'ai' (content generated by AI for each contact — the default when no content is provided) or 'same' (the provided content is sent as-is; supports the [FirstName], [LastName] and [Company] variables). In 'same' mode message is required (max 1900 chars for LinkedIn messages) and email steps also require subject. - invitationNote requires note (max 180 chars, 280 with a Sales Navigator seat). - voiceMessage requires url pointing to the audio file. - likePosts accepts numberOfPosts (1-3, default 1). - visitProfile and invitation need no extra fields. Seats (senders): - Any LinkedIn step → linkedinSeatId is REQUIRED (and not allowed on email-only campaigns). - Any email step → at least one emailSeatId is REQUIRED. Multiple email seats (up to 100) are only allowed when the campaign has email steps EXCLUSIVELY. - There is no seat-listing endpoint; seat IDs can be found on the user's existing campaigns (list_campaigns / get_campaign). When the campaign has a Connection Request step, 1st-degree connections are excluded from the campaign by default.

create_contact

ChatGPT
Create a new contact in Gojiberry. firstName, lastName, and profileUrl are required.

create_list

ChatGPT
Create a new list. Pass campaignId to link the list to a campaign so contacts added to this list are enrolled in that campaign. A list is the link between a source agent (which imports the leads it finds into the list) and a campaign agent (which reaches out to the list's contacts). When setting up a Full Cycle agent (leads sourced and outreached automatically, end to end), create the list FIRST, then the source agent (create_agent, passing this list's ID as listId), then the campaign agent (create_campaign, passing this list's ID in listIds). A list on its own does nothing: it needs a source agent to fill it and a campaign agent to contact what lands in it. Creating a list for a source agent without also creating that campaign leaves the sourced leads sitting in the list, never contacted.

delete_campaign

ChatGPT
Permanently delete a campaign (campaign agent). CANNOT be undone. It leaves the daily outreach scheduling, so no further steps are sent; messages and invitations already sent are not retracted. Its lists and contacts are kept, and the contacts' campaign progress is reset so they can be enrolled in another campaign. Its outreach logs are no longer retrievable with get_campaign_logs. An active campaign can be deleted directly. When the user only wants to stop outreach, use pause_campaign instead. Deleting the campaign of a Full Cycle agent leaves its source filling a list nobody contacts (a 'legacy agent'): mention it, and offer delete_source (or update_source with paused: true) if the user wants the whole agent gone. Ask the user for explicit confirmation before calling this, and show which campaign will be deleted. A missing campaign returns 404.

delete_contact

ChatGPT
Permanently delete a contact. This is not reject_contact (which only removes the lead from the review queue and is reversible with unreject_contact) and not remove_contacts_from_list (which only takes the contact off a list and its campaign). Deleting also stores the contact's LinkedIn profile URL, and its LinkedIn member ID and email when known, in the account's excluded contacts. Sources will not import that person again, and create_contact for the same person is refused (CONTACT_ALREADY_EXCLUDED) until every exclusion covering them is removed: find them with list_excluded_contacts, then delete_excluded_contact each of them. Ask the user for explicit confirmation before calling this, and show which contact will be deleted. A missing contact returns 404.

delete_excluded_contact

ChatGPT
Remove an exclusion so the person can be imported again: sources may find and import them again, and create_contact accepts them. id is the exclusion ID returned by list_excluded_contacts, not a contact ID. This does NOT restore the deleted contact or its history. Ask the user for explicit confirmation before calling this. A missing exclusion returns 404.

delete_source

ChatGPT
Permanently delete a source agent. CANNOT be undone. The source stops sourcing immediately and its run logs (get_agent_logs) are deleted with it. The list it imports into and the leads already imported are NOT deleted. For website-visitor sources, the website tracking script is deactivated. To stop a source temporarily, use update_source with paused: true instead. The campaign that reaches out to the source's list keeps running on the leads already there. Ask the user for explicit confirmation before calling this, and show which source will be deleted. A missing source returns 404.

download_unibox_attachment

ChatGPT
Download an attachment of a unibox message. The message ID and attachment ID come from the messages returned by get_unibox_thread_messages. Images are returned as an image; other files as base64 with their MIME type. Each call fetches the attachment live from the connected LinkedIn account: download each attachment once and reuse it rather than calling this repeatedly. Rate limited to 10 requests per minute.

enrich_contact_email

ChatGPT
Start an email enrichment for a contact. The enrichment runs in the background and this tool returns immediately (the providers can take from a few seconds up to several minutes). To check completion, poll get_contact: the enrichment is done once emailEnriched is true — the email field is then filled if an email was found, and stays empty otherwise. COSTS CREDITS: 1 credit when an email is found; failed enrichments consume nothing.

enrich_contact_phone

ChatGPT
Start a mobile phone enrichment for a contact. The enrichment runs in the background and this tool returns immediately (the providers can take from a few seconds up to several minutes). To check completion, poll get_contact: the enrichment is done once phoneEnriched is true — the phone field is then filled if a phone was found, and stays empty otherwise. COSTS CREDITS: 10 credits when a phone number is found; failed enrichments consume nothing.

get_agent

ChatGPT
Get one source agent's full configuration, including the ICP (target job titles, industries, company sizes, locations, company types, ignored companies, mandatory keywords), its signals (variables — call this before editing them with update_agent, which replaces the whole list) and the listId of the list it imports sourced leads into. Source agents find leads; outreach is done by the campaign agent that uses the same list (see get_campaign) — source agent + list + campaign agent is a Full Cycle agent. To know which campaign outreaches this agent's leads, look its listId up by ID with get_list and read the list's campaignId. An empty campaignId (null or absent) means the list is linked to no campaign, so the leads this agent sources are never contacted: tell the user and offer to add the campaign with create_campaign.

get_agent_logs

ChatGPT
Get up to 100 recent run logs for a source agent, newest first. Useful for questions about how many leads an agent created, which signals produce leads, or whether recent signal runs failed.

get_campaign

ChatGPT
Get detailed information about a single campaign by ID, including its steps array. Campaign membership is driven by list membership, but the link is stored on the LIST: the campaign returned here carries no list reference, so use list_lists and keep the lists whose campaignId matches this campaign's id to know who it reaches out to. Engagement metrics (reply rate, send rate, connection acceptance rate, etc.) are derived from the contacts in the campaign's list: each contact carries a campaignStatus array of per-step records, each with a type (e.g. message, voiceMessage, email, connection, view, ...) and a state (e.g. sent, delivered, answered, skipped, ...). A contact counts as a reply when at least one entry has a type of message, voiceMessage, or email AND a state of answered; the reply rate is the number of contacts with a reply divided by the total contacts in the list. Other metrics follow the same pattern with different type/state combinations.

get_campaign_email_queue

ChatGPT
Preview the upcoming emails of an email-only campaign (every step is an email, no LinkedIn seat), as shown in the Scheduled tab of the app. For LinkedIn and multi-channel campaigns use get_campaign_queue instead — this tool fails on them. Returns one entry per scheduled contact: the contact, the campaign, the campaignStep email to send with its subject and body filled with the contact variables, lastContactStep (the last step the contact already received, absent for a contact entering the campaign), emailSeatId (the mailbox that will send it, see list_email_seats) and sendAsNewThread (true when the email starts a new thread instead of replying to the previous one). The queue is capped at the remaining daily sending quota of the campaign's email seats. Preview only: nothing is sent, no new contacts are pulled in, no AI messages are generated and no credits are used. An inactive campaign has an empty queue.

get_campaign_logs

ChatGPT
Retrieve the outreach event log of a campaign (paginated, most recent first). Each entry records an action performed on a contact — e.g. invitation, invitationNote, invitationAccepted, message, voiceMessage, visitProfile, likePosts, email, replied, email-reply, interested, not-interested — with the contact, the campaign ID, and a timestamp. Useful for questions like 'what did this campaign do recently' or 'who accepted an invitation'.

get_campaign_queue

ChatGPT
Preview the upcoming actions of a LinkedIn or multi-channel campaign (one with a LinkedIn seat), as shown in the Scheduled tab of the app. For email-only campaigns use get_campaign_email_queue instead — this tool fails on them. Returns one entry per scheduled contact: the contact, the campaign, the campaignStep to run next (invitation, invitationNote, message, voiceMessage, visitProfile, likePosts or email) with its message filled with the contact variables, and lastContactStep, the last step the contact already received (absent for a contact entering the campaign). The queue is capped at the LinkedIn seat's full daily quotas (shared between the campaigns on that seat; today's actions are not subtracted) and at the remaining daily quota of the email seats. Preview only: nothing is sent, no new contacts are pulled in, no AI messages are generated and no credits are used. An inactive campaign has an empty queue. Useful for questions like 'what will this campaign send next' or 'who is about to be contacted'.

get_contact

ChatGPT
Get detailed information about a single contact by ID.

get_contact_company

ChatGPT
Get the company of one of your contacts, resolved by the API from the contact's LinkedIn company page (companyUrl). Returns the company id (LinkedIn company id), name, linkedinIdentifier, logo, description, website, size, location, industry, youtubeUrl, and crunchbaseProfileUrl. This is the contact's own company, not a directory reveal: it reveals nothing from the shared companies pool and is not a substitute for reveal_companies. Check get_contact first: if the contact already carries its company data (company, companyUrl, industry, companySize, companyLocation, ...), use that instead of calling this tool. Returns 404 when the contact does not exist OR when no company could be found for it (the contact has no LinkedIn company page, or the company is unknown), so a 404 means there is no company to report — do not retry it. Rate limited to 30 requests per minute and 5000 per day.

get_current_user

ChatGPT
Get the profile of the user the calls run as: ID, email, name, organization ID and role, language, country, timezone. When impersonating (impersonate_user or actAsUserId), the impersonated user is returned — use this to confirm whose account you are acting on.

get_directory_industries

ChatGPT
Free. Returns the suggested values for the industry filter of search_leads and search_companies, both as a flat list and grouped by category. Other industry values are accepted by the search tools, but these labels line up with the directory taxonomy and give the best results.

get_directory_locations

ChatGPT
Free. Returns the suggested values for the location filter of search_leads, grouped by country: the country name itself plus one entry per region or state, formatted as '<region>, <country>' (e.g. 'California, United States'). Other location strings are accepted by search_leads, but these values match how the directory spells locations and give the best results.

get_intent_type_counts

ChatGPT
Get contact counts grouped by intent type. Useful for understanding the distribution of contact intents.

get_list

ChatGPT
Get a single list by ID, including its contacts and its campaignId — the campaign agent that reaches out to them. Use this to resolve a source agent's listId into the campaign behind it: an empty campaignId (null or absent) means the list is linked to no campaign, so its contacts receive no outreach.

get_organization

ChatGPT
Get the organization of the authenticated user (company name, website, industry, size, owner ID, etc.).

get_team_activity

ChatGPT
Team totals for the signed-in user's own workspace over a UTC date range, owner included: the same figures as the Team activity page. Pass both dateFrom and dateTo (YYYY-MM-DD, UTC calendar days; dateFrom may equal dateTo). Convert "this week", "last month", and similar into those dates yourself — never call without both. Today is not in the data: each UTC day is snapshotted at 04:30 UTC the next morning. If the user asks for today, or a range that includes today, say the figures stop at yesterday and use yesterday as dateTo. Only organization owners and admins can see this, for their own workspace; a member is refused. There is no organization id to pass. The call always runs as the API key owner: it does not send x-impersonate-user-id, ignores session impersonation, and has no user id argument. Do not pass an account id to impersonate_user or actAsUserId. A 403 means only owners and admins can see team activity — tell the user that and do not retry with another user id. A 404 means the API deploy has not reached production yet — say that and do not guess a different path. summary: leadsFound, linkedinInvitationsSent, replied, messagesSent, replyRate, interested, pipelineGenerated. Each metric is { value, rate, changePercent }. rate is a ratio from 0 to 1, or null: 0.12 means 12%, so say it as a percent. Acceptance rate (linkedinInvitationsAccepted.rate) = accepted ÷ sent. Reply rate = replied ÷ messages sent. Interested rate = interested ÷ replied. Messages sent = LinkedIn messages (InMail and voice included) + emails. A sent InMail counts as a LinkedIn message and as a contacted lead. changePercent compares with the previous period of the same length. It is null when that previous period is missing a day's snapshot, or when the previous value is 0. Do not invent a trend from null. Interested rate is also null when the period mixes two historical ways that count was stored. leadsFound, linkedinInvitationsSent, messagesSent, and pipelineGenerated have no rate. summary.replyRate.value is the reply-rate ratio (replied ÷ messages sent), the same number as summary.replyRate.rate and summary.replied.rate, not a count. pipelineGenerated.value = replies × average deal size, and is null when no usable average deal size is set. accountsCount is every active member, including those with no activity in the period. dealSizeSetCount is how many members set their own average deal size. generatedAt is the ISO datetime of the newest snapshot in the period, or null.

get_unibox_for_contact

ChatGPT
Legacy endpoint: get the LinkedIn (and email) conversations the user has had with a specific contact. Returns chats with their messages, ordered by most recent message first. The same data is available with pagination through list_unibox_threads and get_unibox_thread_messages.

get_unibox_thread_messages

ChatGPT
Get paginated unibox messages for a thread owned by the authenticated user.

get_user_permissions

ChatGPT
Get the organization permission keys granted to the user the calls run as (e.g. organization.members.impersonate, organization.billing.manage). When impersonating, the impersonated user's permissions are returned. Use it to check whether an organization action is allowed before attempting it.

impersonate_user

ChatGPT
Act on behalf of another member of the organization. Once set, all subsequent tool calls in this session (contacts, campaigns, lists, agents, unibox, ...) are executed as that user, until impersonation is changed or cleared. Only works when the current API key belongs to the ORGANIZATION OWNER, and the target user must be a member of the same organization (member IDs are returned by list_organization_members). Passing userId: null stops impersonating and restores the key owner. This stores the impersonation on the current MCP session. For clients that open a new MCP session per tool call, the stateless actAsUserId argument available on every tool provides the same behavior on a per-call basis.

list_agents

ChatGPT
List all source agents. A source agent is a lead finder: it continuously sources prospects matching the user's ICP and scores them with AI, importing them into its list. It does NOT do outreach — outreach (LinkedIn + email) is done by the campaign agent (see list_campaigns). Every source agent is linked to exactly one list (listId), and that same list is what a campaign agent reaches out to: list + source agent + campaign agent together form a Full Cycle agent. A source agent whose list is not used by any campaign is a 'legacy agent': it only sources leads, with no outreach. To find out which campaign (if any) outreaches an agent's leads, take the agent's listId and look that list up by ID with get_list: the list's campaignId is the campaign doing the outreach. When that campaignId is empty (null or absent), the list is linked to no campaign and the agent's leads are never contacted — offer to complete the setup with create_campaign. To check every agent at once, list_lists returns the campaignId of all lists in one call.

list_campaigns

ChatGPT
Retrieve campaigns. A campaign is the campaign agent — what the Gojiberry platform calls an 'Agent': it performs the outreach (LinkedIn invitations, messages, emails, ...) to the contacts of its lists. Lead sourcing is done separately by source agents (see list_agents). By default (when status is omitted) this returns ALL campaigns — active AND inactive. Do not pass status unless the user explicitly asked for only active or only paused campaigns. status values: - "all" (same as omitting the argument): every campaign, regardless of state. - "active": ONLY campaigns currently running. - "inactive": ONLY paused campaigns — NOT all campaigns. If the user has no paused campaigns this returns an empty array even though active campaigns exist. Campaign membership in Gojiberry is driven by list membership, but the link is stored on the LIST, not on the campaign: the campaign objects returned here carry no list reference. To find which lists feed a campaign, call list_lists and keep the lists whose campaignId matches the campaign's id. Engagement metrics (reply rate, answered count, etc.) are derived from the contacts in the campaign's list: each contact carries a campaignStatus array of per-step records, each with a type (e.g. message, voiceMessage, email, connection, view, ...) and a state (e.g. sent, delivered, answered, skipped, ...). A contact counts as a reply when at least one campaignStatus entry has a type of message, voiceMessage, or email AND a state of answered; the reply rate is the number of contacts with a reply divided by the total contacts in the list.

list_contacts

ChatGPT
Retrieve contacts with optional pagination, search, and filtering. Returns a paginated list of contacts.

list_email_seats

ChatGPT
List the email seats (connected Gmail / Outlook mailboxes) of the authenticated user. Each seat has an id (usable in the emailSeatIds field of create_campaign), the mailbox email, sender name / firstName / lastName, type ('google' or 'outlook'), sending limits (dailyEmailMax, dailyEmailCount, launchHour, activeDays), tracking options (trackOpening, removeUnsubscribeLink), and warm-up status (warmupEnabled, warmupCompleted, warmupStartedAt). Call this before create_campaign when the campaign has email steps, to pick the seat(s) that will send them. When impersonating, the seats of the impersonated user are returned.

list_excluded_contacts

ChatGPT
List excluded contacts: people the sources must never import again. A contact is excluded when it is deleted (in the platform or with delete_contact); its LinkedIn profile URL, LinkedIn member ID and email, when known, are kept here. Sources skip any lead matching an exclusion, and create_contact refuses it (CONTACT_ALREADY_EXCLUDED). Use this to find the exclusions of a person, then delete_excluded_contact each of them (by id) to allow the person to be imported again. A person can be covered by several exclusions (e.g. one holding their profile URL, another their email) and stays blocked while any of them remains: pass every identifier you know (identifier, profileId, email) to find them all. When several filters are given, an exclusion matching ANY of them is returned. Most recent exclusion first.

list_lists

ChatGPT
Retrieve all lists with their contact counts. Each list carries a campaignId: the campaign agent that reaches out to its contacts, and the only place that link is stored. When campaignId is empty (null or absent), the list is linked to no campaign and its contacts get no outreach — if a source agent imports leads into it (see list_agents), those leads are sourced but never contacted, and the Full Cycle setup (list + source agent + campaign agent) is incomplete.

list_organization_members

ChatGPT
List all members of the authenticated user's organization, with their user ID, name, email, role, and active status. When the API key belongs to the organization owner, a member's id is a valid target for impersonation (impersonate_user or the actAsUserId argument).

list_sources

ChatGPT
List all source agents. A source agent is a lead finder: it continuously sources prospects matching the user's ICP and scores them with AI, importing them into its list. It does NOT do outreach — outreach (LinkedIn + email) is done by the campaign agent (see list_campaigns). Every source agent is linked to exactly one list (listId), and that same list is what a campaign agent reaches out to: list + source agent + campaign agent together form a Full Cycle agent. A source agent whose list is not used by any campaign is a 'legacy agent': it only sources leads, with no outreach. To find out which campaign (if any) outreaches an agent's leads, take the agent's listId and look that list up by ID with get_list: the list's campaignId is the campaign doing the outreach. When that campaignId is empty (null or absent), the list is linked to no campaign and the agent's leads are never contacted — offer to complete the setup with create_campaign. To check every agent at once, list_lists returns the campaignId of all lists in one call.

list_team_activity_members

ChatGPT
One row per active member of the signed-in user's workspace, owner included, plus each member's campaigns: the same figures as the Team activity page. Pass both dateFrom and dateTo (YYYY-MM-DD, UTC calendar days; dateFrom may equal dateTo). Convert "this week", "last month", and similar into those dates yourself — never call without both. Today is not in the data: each UTC day is snapshotted at 04:30 UTC the next morning. If the user asks for today, or a range that includes today, say the figures stop at yesterday and use yesterday as dateTo. Only organization owners and admins can see this, for their own workspace; a member is refused. There is no organization id to pass. The call always runs as the API key owner: it does not send x-impersonate-user-id, ignores session impersonation, and has no user id argument. Do not pass an account id to impersonate_user or actAsUserId. A 403 means only owners and admins can see team activity — tell the user that and do not retry with another user id. A 404 means the API deploy has not reached production yet — say that and do not guess a different path. Follows pagination while hasMore is true, up to 500 members, 10 requests, or a response size limit. pagination.total is the count of active members. When pagination.hasMore is true, call again with pagination.nextOffset before telling the user the list is complete. When it is false, no members remain after this response: if you started at offset 0 it is the whole list, otherwise combine it with the earlier responses. Optional limit (1–100, default 100) is the page size of each request, not how many members come back. Optional offset (default 0) is where to start — pass it only to continue a previous response. sortBy: leadsFound, contacted, linkedinInvitationsAccepted, messagesSent, replied, or interested. sortDirection: asc or desc, default desc, and only applied when sortBy is set. Without sortBy, members come back ordered by name and sortDirection is ignored. Each member: accountId, firstName, lastName, email, campaignsCount, activeCampaignsCount, outreach { linkedin, email } (channels used during the period), metrics, campaigns[]. accountId identifies the member and is informational only. Each campaign: campaignId, accountId, name, status (active, paused, leads_sourcing_only, or outreach_only), outreach, metrics. Member and campaign metrics: leadsFound, contacted, linkedinInvitationsSent, linkedinInvitationsAccepted, messagesSent, replied, interested. Each metric is { value, rate, changePercent }. rate is a ratio from 0 to 1, or null: 0.12 means 12%, so say it as a percent. Acceptance rate (linkedinInvitationsAccepted.rate) = accepted ÷ sent. Reply rate = replied ÷ messages sent. Interested rate = interested ÷ replied. Messages sent = LinkedIn messages (InMail and voice included) + emails. A sent InMail counts as a LinkedIn message and as a contacted lead. changePercent compares with the previous period of the same length. It is null when that previous period is missing a day's snapshot, or when the previous value is 0. Do not invent a trend from null. Interested rate is also null when the period mixes two historical ways that count was stored. leadsFound, contacted, linkedinInvitationsSent, and messagesSent have no rate.

list_unibox_threads

ChatGPT
Get paginated unibox (unified inbox) threads. Provide seatId to fetch threads for a specific LinkedIn seat; omit it to fetch threads across all seats in the account.

pause_campaign

ChatGPT
Deactivate (pause) a campaign: it leaves the daily outreach scheduling and stops sending new steps. Messages and invitations already sent are NOT retracted, and the campaign, its lists and the contacts' progress are all kept — reactivate it with activate_campaign to resume where it left off. This is the safe way to stop a campaign; use it instead of deleting when the user asks to stop, pause, or halt outreach. Returns the resulting campaign state, so active on the response confirms the change. Calling this on an already-paused campaign is a no-op. Rate limited to 10 status changes per minute across activate_campaign and pause_campaign, so pause campaigns one at a time and expect a rate-limit error when stopping more than 10 at once.

reject_contact

ChatGPT
Reject a lead: the contact disappears from the review queue, is excluded from campaign scheduling (readyForCampaign is reset to false), and will not be picked up by automatic campaign launches. The rejection records who rejected the lead and when (rejectedAt / rejectedByUserId on the returned contact). Providing a reason is optional but recommended: reasons are stored on the contact and help understand why leads are rejected so targeting can be refined. Use one of the stable reason codes when the lead does not match the targeting; use 'other' together with reasonText for anything else (reasonText is required with 'other' and ignored for the other codes). Calling this on an already-rejected contact only updates the reason (latest reason wins); the original rejection timestamp and reviewer are kept. A rejection is reversible with unreject_contact.

remove_contacts_from_list

ChatGPT
Remove contacts from a list. Because campaign membership is driven by list membership, this also removes the contacts from any campaign linked to the list. The contacts themselves are not deleted.

reveal_companies

ChatGPT
Reveal full company objects from the directory (name, description, website, LinkedIn URL, funding data, exact headcount, ...). This is intended for detailed company research beyond the company information included when revealing a lead. Never call it as a follow-up to reveal_leads just to get the leads' companies; that would use the user's credits a second time for data they already have. Revealed companies are returned in the tool response only. They are NOT imported or saved into the Gojiberry account. Credit usage: 0.5 credit from the user's Gojiberry plan per company actually returned (ids that resolve to no document use nothing). Revealing the same id twice uses credits twice, so avoid re-revealing ids you already have. If the remaining credits cannot cover all requested ids, only the first ids that fit are revealed. User confirmation required: never call this tool on your own initiative. Before calling, tell the user how many companies would be revealed and how many credits that uses, and wait for them to confirm. A generic request like 'find me companies' is not a confirmation — only reveal once the user has approved the credit usage. ids come from search_companies (the id field — the numeric LinkedIn organisation id). Max 100 ids per request; duplicates are de-duplicated.

reveal_leads

ChatGPT
Reveal full lead objects from the directory (name, company, LinkedIn profile URL, website, country, ...). Emails/phones are NOT included — enrichment is a separate feature. COSTS CREDITS: 0.5 credit per lead actually returned (ids that resolve to no document are not charged). Revealing the same id twice charges twice, so avoid re-revealing ids you already have. If the balance cannot cover all requested ids, only the first affordable ids are revealed and charged. REQUIRES EXPLICIT USER CONSENT: never call this tool on your own initiative. Before calling, tell the user how many leads would be revealed and the credit cost, and wait for their explicit confirmation. A generic request like 'find me leads' is NOT consent to spend credits — only reveal after the user has approved the spend. Pass listId (one of your lists, see list_lists) to also save the revealed leads as contacts into that list at no extra credit cost — leads already in the workspace, excluded, or missing a first/last name are skipped. The save is synchronous, so a large reveal with listId can take noticeably longer. ids come from search_leads (the id field — the LinkedIn member id). Max 100 ids per request; duplicates are de-duplicated.

search_companies

ChatGPT
Search the Gojiberry companies directory (shared pool). Free — uses no credits. Returns MASKED company previews: only id, size, industry, and headquarters are filled; every other field (name, website, description, ...) is returned empty. Use reveal_companies with the returned ids to get the full objects (0.5 credit each). Use this company directory only when the user explicitly wants detailed company data (description, funding, exact headcount, ...). It is NOT needed to get a lead's company: reveal_leads already includes it with no extra credits. Companies searched or revealed here are not imported into the Gojiberry account. At least one filter is required. Results are deterministic (same filters + page always return the same companies). technologiesUsed disclaimer: a company 'uses' a technology when that technology was mentioned in at least one of the company's job offers in the past months. It signals the company hires for that technology, NOT a verified install base.

search_leads

ChatGPT
Search the Gojiberry leads directory (shared pool) to buy leads matching your criteria. FREE — no credits are consumed. Returns MASKED lead previews: only id, firstName, jobTitle, location, and industry are filled; every other field (lastName, company, profileUrl, ...) is returned empty so you can see which columns become available after revealing. Use reveal_leads with the returned ids to get the full objects (0.5 credit each). At least one filter is required (no filters returns an empty result). Results are deterministic: the same filters + page always return the same leads. Deep paging is capped at 10,000 results. For the industry filter, prefer the values returned by get_directory_industries (it's a soft preference/boost, not a hard filter).

send_unibox_message

ChatGPT
Send a social message to a real person in an existing unibox thread (delivered through the thread's social conversation). This delivers an actual message to the thread's attendee and cannot be undone. chatId is the internal thread ID returned as id by list_unibox_threads, not the social conversation URN.

unreject_contact

ChatGPT
Reverse a rejection: clears rejectedAt, rejectedByUserId and the stored reason, putting the contact back in the review queue. This does NOT approve the lead — readyForCampaign is reset to false, so the contact still needs to be accepted (e.g. with update_contact setting readyForCampaign to true) before it is picked up by a campaign. Because readyForCampaign is always reset, calling this on an approved contact moves it back to the review queue.

update_agent

ChatGPT
Update a source agent's configuration, including its ICP (job titles, industries, company sizes, locations, etc.) — the ICP lives on the source agent.

update_campaign

ChatGPT
Update a campaign's name, associated list IDs, and/or the MESSAGES of its steps. This tool CANNOT start or stop a campaign — use activate_campaign and pause_campaign for that. IMPORTANT — this tool edits the messages inside existing steps, NOT the steps themselves: - Steps CANNOT be added, removed, or reordered. If steps is provided, it must contain exactly the same number of steps, in the same order, with the same type values as the existing campaign (the API rejects the update otherwise). The current steps array is returned by get_campaign; only the message fields of that array should be edited. - Only the message content (message, subject, note) and messageMode of each step are applied; other step settings (delays, attachments, etc.) are left unchanged. messageMode (for message and email steps only): - same — every contact receives the same content; the message body is required (message, plus subject for email steps). - ai — no message content needed; Gojiberry generates a contextual, personalized message per contact.

update_contact

ChatGPT
Update an existing contact's fields. Only provided fields will be changed.

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/gojiberry"><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/gojiberry)