archive_target
ChatGPTconfirm: true AND must be confirmed with the user first.By Draftboard
Draftboard finds the warm introduction to the person you want to reach. Ask who can introduce you to a prospect, and you get the people in your own network who know them, how strong each relationship is on a 0-100 scale, and the shared history behind it, such as years at the same company or the same school. Use it to see your best introduction opportunities right now, check whether you are already connected to someone, review every path to one prospect, see which of your saved prospects a teammate can reach, and track how your introductions are going. You can also rate the people you would ask for an introduction, label who they are to you, and save new prospects from LinkedIn. Draftboard works only on the data in your own Draftboard account, and every answer says how much of it was checked. Sign-in happens in your browser and can be revoked at any time from Settings, Connected apps. A Draftboard account with a scanned network is required.
confirm: true AND must be confirmed with the user first.{status, isTarget, targetId, hasPaths, pathsCount, topConnector, topRank, degree, directlyConnected}. status is target, not_a_target, import_pending, or lookup_failed. On import_pending and lookup_failed the booleans are null, NOT false — the answer is unknown, so never tell the user they have no path on the strength of it; re-check instead. By default it imports only the URLs that are not targets yet; Draftboard processes an import batch asynchronously, so those usually come back import_pending on this call and resolve moments later. For a SINGLE person where you only need the id or a yes/no, resolve_target is one call instead of two.list_pool ids) into saved targets — idempotent. Every requested id still in the pool is confirmed; nothing is held back. Returns confirmedCount and remainingCapacity (informational only, not a limit on what gets confirmed). This is how you keep the good people a search found; the rest can be left or reject_pooled. Once confirmed they are real targets and their warm paths are charged like any Target's — tell the customer before confirming. list_targets / find_top_paths will include them (path scoring is async, so paths may take a moment).query) and/or by the labels the company carries (labels, labelsMatch). Returns { companies: [{ id, name, connectorsCount, labels[] }], count, nextPage }. This is where a company id for set_company_labels comes from — id here is NOT an account id from list_accounts. connectorsCount is how many people in the network work there, i.e. how many a company label would reach. One company can show up as several rows (different spellings or records of the same employer), each with its own count — offer them all, do not pick silently. Paginated — loop until nextPage is 0.other_quality_signal, which carries NO work/school overlap label. Describe those only from rankDetails (e.g. "you both know 500 people") and never call them strong. One call, one path per target; for several connectors on one person use get_target_connections instead. Scope with accountId, tagNames, title, roles (job-title words on the target, any of them; expand synonyms yourself, e.g. founder, co-founder, cofounder), ownerIds, and ratings for paths only through people you would actually ask (e.g. [4,5]). Each opportunity carries connectorRating, your own star rating of the connector (higher is better), absent when you have not rated them. Filter by connector label with connectorLabels (+ connectorLabelsMatch all|any) — e.g. ["investor"] for paths through investors — and excludeConnectorLabels; each opportunity carries connectorLabels (org labels are shared with the whole team, personal ones are the caller's own, supporter mirrors a 2..5 rating). do_not_contact is a plain label that hides nobody unless the user asks to exclude it.scoreDetails, plus the team members who can make the ask. Answers 'who can <this person> introduce me to?'. Each item may also carry relationships — how the connector and that target know each other, zero or more of exactly current_colleague, former_colleague, university_classmate — and relationshipDetails, the structured records behind scoreDetails (one per shared company / school / mutual-contact signal, exactly one of employment / education / mutualConnections set per record, dates ISO yyyy-MM-dd). BOTH KEYS ARE ABSENT WHEN EMPTY (never []), so read them as item.relationships ?? []: absence means "we hold no structured signal for this pair", NOT "these two have no relationship" — promote on the signal, never demote on its absence, and keep reading scoreDetails, which carries the human-readable summary (it too is omitted when there is nothing to say). The two fields are independent: a mutual-contacts-only signal yields a relationshipDetails record and NO relationships entry, so never derive or index-align one from the other. The response's connector object also carries your personal star rating (1..5, higher is better; absent when unreviewed).list_network_connections (same names, same meaning — ownerIds from get_me for your own or a teammate's network, omit for the team's combined list). Returns { total, unrated, ratings[{rating,count}] 5..1, labels[{label,scope,count}] (non-zero only), zeroLabels[] }. EACH SECTION IGNORES ITS OWN FILTER: a label's count is what list_network_connections with labels: [that label] plus your other filters would return (your labels/labelsMatch/excludeLabels are not applied to it); ratings and unrated ignore ratings/rating/tiers; total applies every filter and equals that list's count. Ratings are the caller's own stars, 1..5 where HIGHER IS BETTER (5 = 'ask anytime', 1 = 'do not ask', which includes the people it hides); never describe a tier in stars. Personal labels and supporter are the caller's own as well. A label count of 0 means nobody has been given that label yet — labels are only what the team has set — NOT that the network holds no such people. When unrated is above 0, it is worth suggesting the user rate them (set_connector_tier). This only counts: to see the people, call list_network_connections with the same filters.{ id, name, customerProfileId, user{ id, firstName, lastName, linkedinUrl }, teamMembers[]{ id, firstName, lastName, linkedinUrl }, credits }. Call this first to confirm whose account you are working with. credits is the team's credit balance: writes are paid from it and are refused only once it reaches 0. If credits is missing, the balance is unknown, not 0 — go ahead and try the write. customerProfileId is your own member id — pass it as ownerIds to scope to your own network (e.g. list_network_connections for 'my connections'). teamMembers is your team roster, you included — to filter through a teammate, match them by name here and pass their id as ownerIds. If several members fit the name, or none fits exactly, ask the user which one (full names) instead of guessing.score (0-100, rank in the outcome tools), scoreDetails (the human-readable shared history — omitted when there is nothing to say, so read it as scoreDetails ?? []), and the team members (owners) whose network the connection comes from. Each connection also carries rating: YOUR star rating of that connector (1..5, higher is better), absent when you have not rated them — unrated is not a low rating. Filter by ownerIds to see paths through specific teammates, by ratings (e.g. [4,5]) for paths only through people you would actually ask, and by connectorLabels (+ connectorLabelsMatch all|any) / excludeConnectorLabels for paths through, say, investors ("paths to this person through our investors"). Each connection carries labels: who that connector is to the team (org labels are shared with the whole team, personal ones are the caller's own, supporter mirrors a 2..5 rating). do_not_contact is a plain label that hides nobody unless the user asks to exclude it. Paginated. A connection may also carry two structured relationship fields. relationships: how the connector and the target know each other — zero or more of exactly current_colleague, former_colleague, university_classmate, and no other value is ever emitted. relationshipDetails: the machine-readable facts behind scoreDetails — one record per shared company / school / mutual-contact signal, with exactly one of employment (company, department, location, overlapStartDate, overlapEndDate as ISO yyyy-MM-dd, loose, unit) / education (school + the same window) / mutualConnections (count) set, plus that record's score. BOTH KEYS ARE ABSENT WHEN EMPTY — the key is simply not in the JSON, it is never [] — so read them as connection.relationships ?? []. They are present when we hold that signal for the pair and omitted when we do not: absence means "we hold no structured signal for this pair", NOT "these two have no relationship". PROMOTE on the signal, NEVER demote on its absence — never drop or downrank a connector for it; scoreDetails carries the human-readable summary. The two fields are also INDEPENDENT: a connection whose only signal is shared contacts gets a relationshipDetails record and NO relationships entry, and a rank can carry relationships with no records. Never derive, gate or index-align one from the other or from scoreDetails.resolve_target with the same URL: it returns the target as soon as the row exists, before any path. Do NOT confirm with list_targets (by tag, updatedSince or anything else) — it only returns targets that already have a path, so an empty result there is never evidence the import failed. Then poll get_target_connections for the paths.tagNames.query) or by degree. Useful for an account-level view of where warm reach already exists.ownerIds (ids from get_me: your own is customerProfileId, a teammate's is teamMembers[].id) you get exactly the LinkedIn 1st-degree connections of those members, and nobody else; to export the whole team's LinkedIn connections, pass every member's id. Without ownerIds you get the team's combined list, which is broader than LinkedIn: it also holds people added by hand as supporters, colleagues inferred from shared employers, and the team members themselves. Resolve a person by name from get_me yourself; if more than one roster member fits the name, or none fits exactly, ask the user which one, naming the candidates in full, before calling. Find people by name with name (typo-tolerant; if the response carries nameMatch: "approximate" nobody matched exactly, so present the rows as the closest names, not as the person asked for) or by job-title words with roles (expand synonyms yourself, e.g. founder, co-founder, cofounder); query is the older free-text search. Say whose connections you are showing. Each row's owners lists who on the team knows that person (the whole team, not narrowed by ownerIds). At most 100 per page: to read the whole list, keep requesting the next page until nextPage is absent or 0, and say how many you read. SEARCH BY RATING: each row carries the caller's personal star rating: 1..5 where HIGHER IS BETTER (5 = ★★★★★ 'ask anytime' … 1 = ★ 'don't ask'), absent when unreviewed. Each also carries tier, the same setting spelled as the raw wire number: 1..5 where LOWER IS BETTER (tier 1 = 'ask anytime' … tier 5 = 'don't ask'). Filter with ratings — [5] is 'my closest connections'. tiers is the same filter on the wire scale; the two are UNIONED, not intersected. HIDDEN: a connector rated 1 ('don't ask') is hidden from the listing. Without ownerIds that is YOUR rating, and a connector you rated 1 still shows while a teammate keeps them visible, carrying your own rating: 1. With ownerIds it is each listed member's own choice: people a teammate hid are left out of that teammate's list. ratings: [1] lists the ones YOU hid; there is no separate 'hidden' flag. This tool only READS the rating — to SET one, call set_connector_tier. LABELS: each row carries labels (who the person is to the team — org labels are shared with the whole team, personal ones are the caller's own, supporter mirrors a 2..5 rating). Filter with labels (+ labelsMatch all|any) and excludeLabels; do_not_contact is a plain label that hides nobody unless the user asks to exclude it. To SET labels call set_connector_labels.new) awaiting confirm/reject — the people a search found and held for review: everyone search_supporters finds, and those search_accounts found beyond its per-company limit. Each carries id, name, linkedinUrl, headline, accountName, source, tags and icpFits: one entry per supporter search that was given an ideal customer (icp) and found this person, each with campaignId, percent (0-100) and reason. No percent means that search could not assess the person — not a poor fit, and not 0; an empty icpFits means no such search found them. With campaignId, icpFits carries only that search's entry. A person fits when percent is 50 or more — the product's threshold; do not invent others. The pool never hides anyone: every person is returned whatever their fit, unless you pass minIcpFit (0-100, requires campaignId), which keeps only people that search scored at or above it and leaves out those it could not assess. To keep a review short, prefer minIcpFit: 50 over paging through everyone. Review the fits with the customer before confirm_pool: say how many fit and that the rest are a poor fit or could not be assessed, and confirm only the people the customer chooses. Filter by campaignId (the id either search returned) to read one search's results, or by accountId / tagIds / query. Paginated (loop until nextPage is 0). A search fills the pool asynchronously, so an empty result may mean the results are not ready yet — it does not mean the search has finished or found nothing; check back later.id, name, kind (targets, accounts, supporters, extension), status (new, draft, active, paused) and createdAt. The same searches the app's Recent Activity shows; archived ones are not listed. name is how Recent Activity labels the search (e.g. "By person · May 22, 2026"). Show the searches as listed — do not group or count them yourself. Use it to find the search a customer means before pause_search / resume_search, and ask when two names match. Paginated (loop until nextPage is 0).type is either manual (a label the customer created and applied themselves, e.g. via import or attach-tags) or automatic (a system-generated batch marker Draftboard stamps on everything ingested together — usually the date of the campaign/upload/discovery batch, e.g. "20-Apr-2026"). There is no queryable icp tag type. Use to discover tag names/ids before filtering targets. Paginated.maxRank), path count (pathsCount), and tags. Returns only targets that ALREADY have at least one warm-intro path — a saved target still being indexed will not appear here (use resolve_target for those). Filter by tag, status, update time, company (accountId), or title/position (title), name (name, typo-tolerant; when the response carries nameMatch: "approximate" nobody matched exactly, so present the rows as the closest names, not as the person asked for), or role words (roles, e.g. ["founder", "co-founder", "cofounder"] — expand synonyms yourself). To scope to a company, first resolve its name to an id with list_accounts (company search), then pass that id as accountId here — far cheaper than paging the whole target list. To find ONE person by LinkedIn URL, use resolve_target, never a page walk; to find people by name when you have no URL, use name. Paginated — loop pages until nextPage is 0.list_pool ids) — soft-deletes the status-new entries. A rejected person stays rejected: later searches do not bring them back, and there is no way to undo it here — confirm with the customer first. Idempotent; already-confirmed prospects are unaffected. Use to clear out people a search surfaced that you don't want.id for the id-scoped tools), or found: false if they are not one of your targets. Use this for any question about a single named person — "is X already a target", "what's X's target id", "do I have paths to X" (resolve, then get_target_connections). Do NOT page list_targets looking for someone: on a large book that is dozens of slow requests and can stop before reaching them. Unlike list_targets (which returns only targets that already have a warm-intro path), this finds ANY of your non-archived targets — including one imported moments ago whose paths are still being computed, so it is also how you get the id straight after import_targets. found: false is a normal answer, not an error.titles at the given companies — for when you have target companies but not the names yet. companies are domains (acme.com) or LinkedIn company URLs (linkedin.com/company/…); titles form the persona (e.g. "Head of Sales"). Tell the customer before launching: people this search finds are normally added as Targets AUTOMATICALLY, up to a limit per company, and their warm paths are charged like any Target's; only the people beyond that limit wait in the pool for review. Every call launches a new search — do not repeat a call to retry or to check on it. Returns a campaignId; people arrive ASYNCHRONOUSLY with no completion signal. list_targets shows the customer's Targets, new ones included once they have at least one warm path, and resolve_target confirms one specific person; the ones held back are in list_pool (filter by this campaignId), where confirm_pool makes the ones the customer wants Targets (charged the same way) and reject_pool discards the rest.titles inside the networks of the given supporters — people the customer knows and can ask for an introduction — optionally only at the given companies. supporters are LinkedIn profile URLs (linkedin.com/in/…); titles form the persona (e.g. "VP Sales"); companies are domains (acme.com) or LinkedIn company URLs. Launching charges nothing: the people this search finds wait in the pool for review and do not become Targets on their own. Someone the customer rejected earlier stays rejected. Every call launches a new search — do not repeat a call to retry or to check on it. Returns status, errors, a campaignId and counts of supporters accepted and not accepted — not the people found, who arrive over time with no completion signal. Check back later with list_pool filtered by this campaignId, then confirm_pool (a batch of ids) the ones the customer wants and reject_pool the rest. Before confirming, tell the customer that confirming takes those people on as Targets and their warm paths are then charged. The search does not change anyone's rating. Optionally pass icp, the customer's ideal customer (industry, companySize, location, description): every person the search finds is then scored for fit against it. Take it ONLY from what the customer said — never infer, guess or fill in an ideal customer they did not describe. The pool never hides anyone for a poor fit: review the fits with list_pool (this campaignId, optionally minIcpFit) together with the customer. icp cannot be combined with companies — an ideal customer is scored only on a search through the whole network.find_network_companies with the company name and pass the id of the row you pick as companyId. NEVER use an account id from list_accounts or search_accounts — it is a different kind of id and is refused. One company can appear as SEVERAL rows (e.g. a long and a short spelling of the name), each with its own connectorsCount: show the user every row with its count and let them choose one or several, then call this once per chosen id. State how many people the label will reach (connectorsCount) before writing. do_not_contact is only a label and hides nobody by itself. Adding a label already there changes nothing.excluded: true) or un-exclude (excluded: false) a connector for this customer. Excluded connectors are dropped from warm-path results. The product moved this axis onto the rating: a rating: 1 ('don't ask') hides the connector and sets this flag for you, so prefer set_connector_tier with rating: 1. ⚠ The sync runs ONE WAY: writing a rating updates this flag, but excluded: false here does NOT clear a rating: 1 — it leaves the connector rated don't-ask while un-excluded. To un-hide someone cleanly, give them a rating of 2..5 instead of un-excluding.{"add":["investor"]}). ORG labels — investor, customer, advisor, partner, do_not_contact — are visible to the WHOLE team, with who set them; PERSONAL labels — close_friend, family, mentor — are visible only to the caller. supporter is NOT a label write: it is the caller's 2..5 star rating, so for 'make them a supporter' / 'star them' call set_connector_tier. The automatic labels (current_colleague, former_colleague, university_classmate) cannot be set or removed. do_not_contact is only a label: it marks the person and hides nobody by itself. Adding a label already there, or removing one that is not, changes nothing. To label a whole COMPANY (everyone who works there) use set_company_labels. Read a person's labels back on labels of list_network_connections, get_target_connections and find_top_paths.preferred: true) or unmark (preferred: false) a connector as a preferred supporter for this customer. Preferred connectors are prioritized when ranking warm paths. The product moved this axis onto the rating, so reach for set_connector_tier for any new work and use this tool only when the caller specifically needs the preferred flag itself. This flag is NOT the rating: set_connector_tier does not write the preferred column, so setting rating: 5 will not mark someone preferred, and marking them preferred will not give them a rating.rating for a connector — how readily they'd ask this person for an intro. This is the tool for 'star this person', 'rate them' or 'prioritize them as a supporter'. rating is 1..5 and HIGHER IS BETTER: 5 = ★★★★★ 'ask anytime' (the closest) down to 1 = ★ "don't ask", with 4, 3 and 2 in between — and a rating of 1 also HIDES the connector from the default listings. tier is the same setting spelled as the raw wire number, where LOWER IS BETTER (1 = 'ask anytime' … 5 = "don't ask", 0 = clear), and is still accepted: pass EXACTLY ONE of rating (1..5) or tier (0..5) — sending both, or neither, is rejected. There is no rating: 0: clearing a rating back to unreviewed stays tier: 0. This does NOT touch the legacy preferred flag — that is a separate column with its own tool. To SEARCH by rating instead, call list_network_connections (each supporter carries its rating; filter with ratings). Personal to the API-key owner.requested (intro sent), completed (intro made), or declined. For declined you may pass reasonId (connector_declined | prospect_declined | other) and a free-text customReason (≤200 chars). Idempotent per status.Click any prompt to copy it.
Add this badge to your site or README. It links to this page.
HTML
<a href="https://mcp-app-store.keenethics-labs.com/app/draftboard"><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
[](https://mcp-app-store.keenethics-labs.com/app/draftboard)