Skip to content
ProspectAPIs

The ProspectAPIs
changelog

The ProspectAPIs changelog is a dated record of every change you can actually use: new endpoints, price changes, MCP tools and pages, newest first. Each entry carries the day it reached customers rather than the day it was written, so you can tell when something became available to you.

The ProspectAPIs changelog lists every change you can use in the REST API, the MCP server and this site, by the day it shipped. Per our release history, 43 releases shipped between October 1, 2026 and November 12, 2026, and every public route the API serves is named in at least one of them. The latest release is from : resume a paused watchlist, share urns resolve, a tool list that matches what is on sale. The API reports its running version in the X-API-Version header and at GET /v1/version.

Resume a paused watchlist, share URNs resolve, a tool list that matches what is on sale

  • POST /v1/watchlists/:id/resume (MCP watchlist_resume) switches a watchlist back on after it paused itself for repeated webhook failures, keeping its id and delivery history. Free. A watchlist paused for another reason answers 409 not_resumable.
  • post_engagers given a share URN or a /posts/…-share-… URL now resolves to the post and returns its reactions or comments, instead of answering 503.
  • The hosted MCP server lists only the tools your key can call right now, so a tool that is not on sale yet is no longer offered.
  • Connecting an account with only part of a session (a bare cookie value) answers 422 session_incomplete, listed on the connected accounts docs and in the OpenAPI document.

Company lookups wait for the build, clearer people reads, fixes to holds and reminders

  • A first GET /v1/companies/:domain for a company that is not built yet now waits up to 20 seconds for the build and answers with the profile. Only a build that takes longer still answers the free 202 building, and the build queue now runs every 10 seconds instead of every minute.
  • People reads: a member the network does not name comes back with hidden: true and a person_ref. A stand-in name such as "LinkedIn Member" and a stand-in token headline now come back null instead of as text.
  • GET /v1/posts/engagers given a share id or share URL it cannot resolve to a post answers a free 404 not_found, never a billed empty page. A reactions page never runs past the post's own total, and the errors page documents upstream_changed: restart from the first page after the wait.
  • Connecting a session through POST /v1/accounts/session or POST /v1/accounts without a user_agent still succeeds, and the answer now carries warnings: [{ code: "user_agent_missing" }] naming the two fixes: send the browser's exact User-Agent, or sign in with remember: true.
  • People reads served by our read pool now have a per-account limit a minute, sized to the pool's live capacity and never below 5 a minute; GET /v1/pricing shows the current value, and a read past it answers a free 429 people_read_rate_limit. An account that has only used the free signup credit can make 100 people reads a day (a free 429 daily_read_limit after that, reset at 00:00 UTC); any top-up removes the daily cap.
  • A background email verification batch that stops part way now settles its hold at what it charged so far, so no held amount is left on your balance.
  • Checkout reminders: at most one reminder and one follow-up a day or more later for an unpaid checkout, then nothing until you complete a payment.
  • Funding rounds from vcnewsdaily.com are read again: the source's slow pages no longer time out.

Ads library on sale, Google first

  • GET /v1/ads and GET /v1/ads/:id are on sale for the Google ads library: whether a company runs Google ads, by domain, and each live ad. Coverage is worldwide for verified advertisers' ads shown in the last 30 days, live when shown in the last 2.
  • $0.006 per library answered with ads (status ok). no_ads, not_resolved, not_covered and unreadable answers are free. GET /pricing lists the price, the covered libraries and both routes as available.
  • The ads library page, /ads-library-api, shows a sample answer and what each status means.

People reads now $0.0012 each, no daily cap; every write $0.005

  • Reads on your own connected account now $0.0012 each (was $0.01): GET /v1/messages, GET /v1/messages/thread, GET /v1/connections, GET /v1/invitations/received, GET /v1/invitations/sent, GET /v1/people/skills, GET /v1/insights/profile-views, GET /v1/insights/ssi and GET /v1/insights/posts cost the same $0.0012 a read as every other read.
  • No cap on connected accounts per API key (the 50-account limit is gone), and POST /v1/website/contacts takes any max_finds: what fits in the request (10 addresses to verify, 4 finds) is answered now, the rest comes back status queued with a job_id, read from GET /v1/email/verify/jobs/{id}/results and billed per contact found when the job delivers it.
  • Enriched page followers now $0.003 each (was $0.01): GET /v1/pages/followers/enriched charges $0.003 per follower returned with its profile, 20 a page with more by cursor.
  • Every write $0.005. Every write route (messages, chats, invites and withdrawals, follows, visits, posts, comments, reposts, quotes, edits and deletes, reactions, endorsements, accepting or declining an invitation, removing a connection, marking a thread read) now costs one flat $0.005 per action, replacing the per-route prices. GET /pricing reports it as people_writes.price_per_write_usd.
  • People reads now $0.0012 each, no daily cap. Every route in the people reads family (GET /v1/people/profile, POST /v1/people/search, GET /v1/people/posts, GET /v1/posts/engagers, GET /v1/posts/resharers, GET /v1/pages/followers, GET /v1/pages/profile, GET /v1/posts/detail, POST /v1/pages/search, POST /v1/posts/search, GET /v1/pages/employees) drops from $0.012 to $0.0012 per live read, and the per-account daily read limit (the daily_read_limit 429) is gone.
  • Pay as you go with no volume caps: the one limit is the per-key requests-a-minute rate limit, answered as a 429 with Retry-After. POST /v1/email/verify and POST /v1/email/clean take up to 50,000 addresses a call (an input size per request, not a volume cap: send more in further calls or as a background job; a call over it is a 400 TOO_MANY_EMAILS, free); an address our probe pacing cannot check now comes back queued with a job_id, free on that call and billed when the job checks it. GET /v1/email/domain/:domain has no hourly limit, there is no cap on watchlists or on actions queued for a disconnected account, and GET /v1/pages/followers/enriched pages 20 followers at a time.

A light profile read, announced ahead of launch

  • GET /v1/people/profile/lite reads the light version of one person's public profile, fresh: name, location, current company name, profile_ref and the public about when the profile shows them.
  • It is charged only for a filled answer, a name with a location or a company name. A profile that is not found or shows too little is a free 404, and a read that cannot be done right now is a free 503 with Retry-After.
  • It answers 503 and charges nothing until GET /pricing lists people_profile_lite as available; then $0.0005 per filled answer. The full profile stays at GET /v1/people/profile. The MCP server adds people_profile_lite.

Lead and company search, announced ahead of launch

  • POST /v1/leads/search searches people as sales leads by keywords and lead filters: seniority, function, company headcount, years in role and at the company, a recent job change and recent posting. Each lead carries a current role with tenure, network distance and signals. The first 2,500 results per query.
  • POST /v1/leads/companies/search searches companies as sales accounts by headcount, headcount growth, annual revenue, industry, location and hiring now.
  • Both answer 503 and charge nothing until GET /pricing lists lead_search as available; then $0.0035 per result returned, and an empty search is free. The MCP server adds lead_search and lead_company_search.

Your page's followers, each with their profile

  • GET /v1/pages/followers/enriched reads one page of the followers of a page your connected account administers and joins each follower to their full profile (headline, location, current company and title, experience, education, skills), read fresh in the same call. Up to 10 followers a call; page through with the cursor.
  • Every follower says ok, no_reference (no public profile to read), not_found or skipped (with the reason: not_held, daily_limit, pool_busy, time_budget or unavailable), so a follower that was not enriched never passes for an empty profile.
  • The route answers 503 and charges nothing until GET /pricing lists follower_enrichment as available; then $0.01 per follower returned with its profile, and nothing for the list or for a follower that was not enriched. The MCP server adds page_followers_enriched.

An event's speakers, sponsors, agenda and side events

  • POST /v1/events/enrich takes one Luma or Partiful event page (or an id from GET /v1/events) and returns the event with four sections: speakers (name, title, company, profile link when the speaker is a host), sponsors (name, tier, link), the agenda (start, end, title, who is on stage) and side events (the event pages it links, each with its title and start).
  • Every section says ok, not_published (the page lists none) or unreadable (we could not tell), so an empty list never passes for none. Anything read from the event's description is marked as such, with the heading it sat under and a confidence.
  • The route answers 503 and charges nothing until GET /pricing lists events as available; then $0.005 per enriched event, charged only when at least one section is found. The MCP server adds event_enrich.

Find events by city on Luma and Partiful

  • GET /v1/events finds public Luma and Partiful events in one city in one record shape: the platform's own id and url, start and end in UTC and in local time with the offset, venue, Luma's own category, the guest count where the host shows it, hosts with their platform ids, the organiser and the ticket price. Filter by date, category, guest count, host, speaker, title and price; every answer says per platform whether it was read, not covered or unreadable.
  • GET /v1/events/cities lists the 87 cities covered and which platforms cover each one, free. GET /v1/events/:id reads one event fresh from its platform. A watchlist with kind events pushes each new matching event in a city to your webhook as event.discovered.
  • Every event is read from the public pages, never through an account. The routes answer 503 and charge nothing until GET /pricing lists events as available; then $0.0015 per event returned. The MCP server adds events_search, events_cities and event_get.

The contacts a company's own website lists

  • POST /v1/website/contacts takes a domain (or an https URL on it) and reads the company's own site live: team, about, contact and press pages, its structured data and, for a URL, that page's author. It returns the people and addresses the site lists, ranked for your objective (sales, partnership, backlink, guest_post, listicle or pr), each with a role, a role_fit, a stable fingerprint, the source_url it was found on and the snippet as evidence.
  • Published addresses are checked at the mailbox, and the top people with no published address get the same pattern check as POST /v1/email/find, so every contact says how its address was reached: published, pattern_verified, catch_all_guess or none. Pass your own company_id and it comes back in key with the domain.
  • $0.009 per contact returned with a usable address. Names without an address, catch-all guesses, a dead published address and a site with no contacts, one that cannot be read or one whose robots.txt disallows us are free. The MCP server adds website_contacts.

Hiring and tech stack by domain

  • GET /v1/companies/:domain/jobs returns a company's open job postings, read from its own public job board (Greenhouse, Lever, Ashby, Workable, SmartRecruiters, Recruitee) or the job postings on its careers page. Each opening carries its source, the board it came from (ats: provider and board_slug), fetched_at and our own first and last seen times.
  • GET /v1/companies/:domain/technologies returns the technologies the company's site and DNS show, each with the evidence that matched (a script, header, cookie name, meta generator or DNS record) and where it was seen.
  • Both read live when the stored answer is older than max_age_hours, and cost $0.006 per lookup that returns data; an empty answer and an unreadable site are free. They answer 503 until GET /pricing lists hiring_tech as available. The MCP server adds company_jobs and company_technologies.

Hosted connect links and account events

  • POST /v1/accounts/connect-link returns a single-use link on prospectapis.com where your own user connects their account: they sign in there, with any code or check it asks for, or paste a session string, and you never handle their password. Links last one hour by default (5 minutes to 24 hours with expires_on), are stored only as a hash, stop after ten attempts and end when the API key that made them is revoked. DELETE /v1/accounts/connect-link/:link_id revokes one. Both are free.
  • type reconnect with an account_id reconnects that account and keeps its id. name, your own reference, comes back on success_redirect_url and failure_redirect_url with status and account_id.
  • A link with notify_url sends that account's account.connected, account.reconnect_required and account.disconnected events there, signed like watchlist deliveries with the webhook_secret shown once. The MCP server adds account_connect_link and account_connect_link_revoke.

Connect many accounts to one key

  • One API key now holds up to 50 connected accounts, each with its own account_id. POST /v1/accounts connects another account (a session string or a sign-in), GET /v1/accounts lists them, GET /v1/accounts/:account_id reads one and DELETE /v1/accounts/:account_id removes one, its stored session and its queued writes, leaving the others untouched.
  • POST /v1/accounts/:account_id/reconnect connects the same account again and keeps its account_id. POST /v1/accounts/:account_id/checkpoint finishes a sign-in that asked for a code or a check.
  • Every read on a connected account and every write takes account_id. A key with exactly one account may leave it out, so existing integrations keep working; with more than one, a call without it answers 400 account_id_required.
  • Each account acts from its own dedicated connection, with its own browser and its own action limits, and the same account cannot be connected twice under one key. All of these routes are free. The MCP server adds accounts_list, account_add, account_get, account_remove, account_reconnect and account_checkpoint.

Your own network: connections and invitations

  • GET /v1/connections lists your connected account's first-degree connections, newest first, each with name, headline, profile_ref, person_ref and connected_at. GET /v1/invitations/received lists the invitations waiting on your account, each with an invitation_id, the sender and the optional message, and GET /v1/invitations/sent lists the ones you sent, with their status. $0.01 per read.
  • POST /v1/invitations/accept and POST /v1/invitations/decline answer a received invitation by its invitation_id, and POST /v1/connections/remove removes a connection by its person_ref. $0.005 per action. Unfollow a person with POST /v1/people/follow and unfollow: true.
  • Each route is on sale once GET /pricing lists it as available; until then it answers 503 and nothing is charged. The MCP server adds connections_list, invitations_received, invitations_sent, invitation_accept, invitation_decline and connection_remove.

Media posts, post and comment management, endorsements and account insights

  • POST /v1/posts takes optional media: up to 20 images (JPEG, PNG or GIF, 5 MB each), one MP4 video (50 MB) or one PDF document (25 MB), 50 MB in all, each by public https address or inline as base64. The caption is optional with media.
  • Manage your own posts and comments: POST /v1/posts/edit ($0.02), POST /v1/posts/delete ($0.02) and POST /v1/posts/comment/delete ($0.02). Endorse a listed skill with POST /v1/people/endorse ($0.005).
  • Read a person's skills with GET /v1/people/skills, and your own account's GET /v1/insights/profile-views, GET /v1/insights/ssi and GET /v1/insights/posts, each $0.01 a read on your connected account.
  • Each comment in GET /v1/posts/engagers?kind=comments now carries a comment_id, which POST /v1/posts/comment takes as parent_comment to reply and POST /v1/posts/comment/delete takes to remove your own.
  • Each new route opens as GET /pricing lists it in available_routes; until then it answers 503 and nothing is charged.

Read a post in full, search companies and posts, list a page's people

  • GET /v1/posts/detail reads one public post in full: its author, text, date, a media summary and its counts (reactions in total and by type, comments, reshares).
  • POST /v1/pages/search searches company pages by keywords, industry, size and location, and POST /v1/posts/search searches public posts by keywords with a date window and sort. Each reads the first 1,000 results per query.
  • GET /v1/pages/employees lists the people who say they work at a company page, each with name, headline and profile.
  • GET /v1/posts/engagers?kind=comments takes include_replies=true to nest each comment's replies under it, at most 10 a comment. Without it the answer is unchanged.
  • Each is a live read at $0.012 and is served once GET /pricing lists it as available; until then it answers 503 and nothing is charged. The MCP server adds post_get, page_search, post_search and page_employees.

Group conversations, attachments and message reactions

  • POST /v1/chats starts a conversation from your connected account with one person or a group of up to 50, with an optional group title, and answers the new thread_id. $0.03, the price of a message.
  • POST /v1/messages now takes one attachment of up to 20 MB (PDF, Word, PowerPoint, Excel, plain text, JPEG, PNG or GIF), as an https address or base64, and inmail: true with a subject to message someone outside your network. Reading a thread lists each attachment's name, type and size.
  • POST /v1/messages/react adds or removes an emoji on one message, and POST /v1/messages/read marks a thread read, $0.005 each. GET /v1/messages/inmail-credits reports the credits left for messages outside your network, free.
  • Each action rolls out once it is switched on: GET /pricing lists it in people_writes.available_routes, and until then it answers 503 with nothing charged. The MCP server adds chat_start, message_react, thread_mark_read and inmail_credits.

Writes wait while your account is signed out

  • A write sent while your connected account is signed out or needs a reconnect now answers 202 with status queued and a queued_id, instead of failing. It is sent through the same pacing once the account is connected again, and charged only when it is sent. Up to 50 writes wait per account, each for up to 72 hours.
  • GET /v1/writes/queued lists the waiting writes with their state, and DELETE /v1/writes/queued/:id cancels one so it is never sent. Both are free. The MCP server adds writes_queued and write_queued_cancel.
  • Cookies the platform refreshes during a session are now saved, so a refreshed sign-in is not lost.
  • A queued write goes out only from the account it was queued under, disconnecting cancels every waiting write, and a send whose answer was lost is never sent twice (outcome_unknown).

Read your feedback back

  • GET /v1/feedback lists the reports your account sent, newest first with limit and cursor, and GET /v1/feedback/:id reads one. Each shows its status (received, triaged, fixed, wont_fix or needs_info) and the team's response. Both are free, from any key on the account.
  • POST /v1/feedback now stores the report before alerting the team, so it is kept even when the alert fails (status stored). The MCP server adds feedback_list and feedback_get.

Email verification: cache-only checks and retry times

  • POST /v1/email/verify takes depth "cache" (or cache_only: true): it answers only from verdicts already on record, never contacts the mail server and never counts against the 30-minute probe budget. A known verdict is billed like a full check; an address with nothing on record comes back miss, free.
  • A paced or provider_cooldown result now carries retry_after_s, the seconds until that mail provider can be checked again.

Launch events, documented ahead of release

  • GET /v1/launches and GET /v1/launches/:id are documented: company launches as prospecting triggers, with the company domain, the kind of launch, a link to the company's own post and the founder. They are not on sale yet: nothing is charged until the price list shows launch events as available.
  • Watchlists accept kind launches for the same records, and the MCP server lists launch_search and launch_get.

Connect your own account

  • POST /v1/accounts/session connects your own LinkedIn account with its signed-in session string, checked live once; POST /v1/accounts/login connects it by signing in, with a two-factor code, an emailed code or a check a person finishes when the account asks.
  • GET /v1/accounts/session shows what is connected (active or signed out, the handle and name, the dates), never the session itself; DELETE /v1/accounts/session disconnects it and deletes the stored session.
  • The session is kept encrypted, used only for the actions you request, never shown to another customer, and deleted when you disconnect. All four are free.
  • The MCP server gains account_connect, account_login, account_status and account_disconnect, and the steps are on the new connected accounts guide in the docs.

Company pages read live, company profiles by name, and a company's ads

  • GET /v1/ads tells you whether a company runs ads now on the Meta, Google, LinkedIn and TikTok ad libraries, by domain: a one-line summary, a status for each library (ok, no_ads, not_resolved, not_covered or unreadable), how the advertiser was found, and the live ads with text, media, dates, countries, landing URL and status where the library shows them. GET /v1/ads/:id returns one ad in full. $0.006 per library that returns ads; everything else is free.
  • An ads watchlist (POST /v1/watchlists with kind ads) delivers advertiser_started, advertiser_stopped and new_creatives for one company to your webhook, or every ad's own events with level ad.
  • The ads library rolls out once every surface ships it: GET /pricing reports ads.available, and the routes answer 503 until then.
  • GET /v1/pages/profile reads a company page's public fields fresh at request time: name, website, description, industry, company type, employee count and size range, headquarters, founded year, specialities, follower count and affiliated pages. $0.012 per live read; a page that cannot be read is free.
  • GET /v1/companies/:domain also takes a company's exact name. A name shared by several companies answers 300 with each name and domain to choose from, never a guess; a name we hold no rounds for is a 404. Both are free.

See who reshared a post, and how

  • GET /v1/posts/resharers reads who reshared a public post, each resharer enriched, fresh at request time, $0.012 per live read.
  • Every result names the reshare type: repost for a plain reshare, or repost_with_thoughts for a quote, which also carries the resharer's own commentary text.
  • Reshare-with-commentary is the highest-intent engagement signal on a post, and no other provider exposes it.
  • It rolls out with the other live reads: GET /pricing reports people_reads.available, and billing turns on with it.

Actions and messaging, through your own account

  • Act through your own connected account, priced per action with no per-seat or per-account fee: POST /v1/messages (send a DM, $0.03), POST /v1/people/invite ($0.03), POST /v1/people/invite/withdraw ($0.01), POST /v1/people/follow ($0.01), POST /v1/people/visit ($0.01), POST /v1/posts (create a post, $0.05), POST /v1/posts/react ($0.005), POST /v1/posts/comment ($0.02), POST /v1/posts/repost ($0.01) and POST /v1/posts/quote ($0.02).
  • Read your own conversations: GET /v1/messages and GET /v1/messages/thread, $0.01 per read.
  • Connecting your own account is free; you pay only for the actions you run through it.
  • The actions roll out as account supply grows: GET /pricing reports people_writes.available and messaging_reads.available, and billing turns on with each.

Cited company profiles by domain

  • GET /v1/companies/:domain returns one cited company record by domain: name, description, HQ, founded year, industry, employee band, social links, a funding rollup and a careers link, with a source on every field and each marked stated or inferred.
  • A field no source supports is null, never guessed; a domain not built yet returns 202 building (retry shortly) and an unknown domain returns 404, both free.
  • GET /v1/companies/:domain/funding returns the funding rounds for one domain, the same records as GET /v1/funding filtered by domain, at $0.009 per record returned.
  • The profile rolls out as coverage grows: GET /pricing reports companies.available at $0.02 per company returned, and billing turns on with it.

Live people and page reads, at one published price

  • Five live reads join the API at $0.012 per read, each fetching the person or page fresh at request time rather than from a stored cache.
  • GET /v1/people/profile reads one person's public professional profile by a public identifier or profile address; a profile that cannot be read is free.
  • POST /v1/people/search searches people by keywords and filters and returns the first 1,000 results per query.
  • GET /v1/people/posts reads a person's recent public posts, and GET /v1/posts/engagers reads who reacted to or commented on a public post.
  • GET /v1/pages/followers reads the followers of a page your connected account administers.
  • The reads roll out as live-account supply grows: GET /pricing reports people_reads.available, and billing turns on with it.

Lower prices on funding records, email checks and found addresses

  • GET /v1/funding and GET /v1/funding/:id now cost $0.009 per record returned ($9 per 1,000), down from $0.02. A search that matches nothing is still free.
  • Watchlist deliveries cost $0.009 per delivered record, the same as a funding record.
  • POST /v1/email/verify costs $0.003 per valid or invalid result ($3 per 1,000), down from $0.004; catch-all and unknown results stay free, and verification jobs bill at the same price.
  • POST /v1/email/find costs $0.009 per found address ($9 per 1,000), down from $0.02; nothing found, a catch-all domain or an unknown answer stays free.
  • The $1 signup balance now covers about 111 funding records or 333 email checks.

Worked examples, a Crunchbase API comparison, and two new ways in for agents

  • /case-studies shows three real calls (a funding search, an email verification and a research brief), each with the date it ran, the request, a trimmed copy of the response and its cost at list prices.
  • /alternatives/crunchbase-api compares the funding API with the Crunchbase API row by row, quoting only Crunchbase's own pages for their side.
  • POST /mcp/free lets an agent try the MCP tools without an API key: 10 calls an hour and 25 a day per address, and 60 a day per network, and each answer says how to get a key for more.
  • Clients that support remote MCP servers can connect to https://api.prospectapis.com/mcp with OAuth: they find the sign-in server through /.well-known/oauth-protected-resource, and the user approves the app instead of pasting an API key.
  • POST /v1/email/verify/jobs accepts {email, first_name, last_name} items like POST /v1/email/verify, so a row whose address follows the person's name gets the name_match flag.
  • Sign-up confirmation, sign-in link and password reset emails now come with a plain text version as well as HTML.

The JavaScript SDK is on npm, and clearer account emails

  • npm install prospectapis installs the JavaScript SDK (0.1.1); the one-file download stays available for the JavaScript and Python clients.
  • The weekly watchlist summary and the monthly usage summary each have one button to the dashboard, and every billing, usage and onboarding email from us says that a person reads your replies.

Bonus credit on larger top-ups

  • A top-up of $100 or more adds 25% bonus credit, $500 or more adds 60%, and $2,000 or more adds 120%, by card or in crypto. The price of every call is unchanged; the bonus grows the balance it draws from.
  • GET /pricing lists the tiers as topup_bonus_tiers, and the billing page shows the credit each amount buys before you pay.

Email verification endpoint, and a $1 free balance replaces the monthly free records

  • POST /v1/email/verify checks 1 to 25 email addresses a call. Each result is valid, invalid, catch-all or unknown, with a reason, flags and the mail server's reply code. A valid or invalid result costs $0.004; catch-all and unknown results are free.
  • When the balance covers only some addresses, the call verifies what it can pay for and returns the rest as unknown, free, with X-Result-Capped: balance.
  • Every account now starts with a free $1 balance, no card needed, and every paid call draws from it at its own price: about 50 funding records, 250 email checks or 6 research briefs. Existing accounts get the same $1. The 100 free funding records a month are gone, and nothing resets monthly.
  • Watchlist deliveries cost $0.02 per delivered record from the balance. GET /pricing adds signup_credit_usd and email_verification; GET /v1/account adds signup_credit_usd and price_per_verification_usd.
  • The MCP server (prospectapis-mcp 0.3.0) adds the email_verify tool.
  • Every email verification result now carries a score from 0 to 100, a recommendation (send, send_with_caution, retry_later or do_not_send) and the reason in plain words.
  • A definite result from a recent check of the same address (30 days for valid, 90 for invalid) comes back with the cached flag and cache_age_days, billed like a fresh check.
  • POST /v1/email/outcomes reports bounced or delivered for up to 100 addresses, free. A bounce you report makes your later checks of that address come back invalid; a delivered report on a catch-all address raises its score.
  • GET /v1/email/verify/stats returns your own verification mix over 1 to 90 days, free. The MCP server adds email_outcomes_report and email_verify_stats.
  • POST /v1/email/verify/jobs checks up to 50,000 addresses in the background. A greylisting or slow server is tried again about 5, 15 and 30 minutes later, free; each valid or invalid result costs the same $0.004 as it is checked; GET /v1/email/verify/jobs/:id reports progress, GET /v1/email/verify/jobs/:id/results returns the results 1,000 a page or as CSV, and a signed callback fires on completion. The MCP server adds email_verify_job_create, email_verify_job_get and email_verify_job_results.
  • Each result also carries did_you_mean for a likely domain typo, mail_provider and checked_at, and flags a full mailbox, a system or machine-made address, a parked or newly registered domain, and a match with the person's name when you send first_name and last_name.
  • POST /v1/email/verify takes depth quick: no mail server is contacted (DNS, lists and recent results only), $0.001 per address whatever the status.
  • POST /v1/email/find returns a person's work address from first name, last name and domain, $0.02 only when one is found. GET /v1/email/domain/:domain answers what a domain's DNS and our recent checks show, free, 300 lookups an hour. GET /v1/email/verify/calibration publishes the measured bounce rate per score band, free.
  • Bounce guarantee: a valid result you were charged for and report bounced within 30 days through POST /v1/email/outcomes is credited back automatically. Credits never expire.
  • POST /v1/email/clean sorts up to 1,000 addresses into keep, review and remove with quick checks. A verification job can check a random sample of a list and estimate the rest, or monitor a list and check it again every 7 to 90 days, reporting what changed; POST /v1/email/verify/jobs/:id/stop turns monitoring off. Job results also stream as NDJSON.
  • Any POST takes an Idempotency-Key header: a retry within 24 hours gets the first answer back and is never charged twice.
  • Keys can be made read only, given a monthly spend cap, or turned into a form-widget key that runs quick checks only from the sites you list; the form widget, a Google Sheets script and JavaScript and Python clients are linked from the email verification docs. The MCP server adds email_find, email_domain_check and email_verify_calibration.

Clearer errors and a faster first call

  • A round filter takes common spellings (SeriesA, Series A, preseed) and refuses a value no round can match with a 400 that names valid stages, instead of an empty answer.
  • Sending the docs' sample key as it is (pa_live_...) gets a 401 that says so, and a 402 now links straight to Billing.
  • The dashboard usage list shows every successful free call, with the time zone on each row.
  • The account confirmation email links to prospectapis.com, and the quickstart's samples run as pasted: JavaScript, Python and three complete sample apps.

Try the MCP server with no key

  • Any MCP client can now connect to POST /mcp/free at https://api.prospectapis.com/mcp/free with no key and no account: funding search (the first five rounds), one funding record by id, the price list and the API version.
  • The free tier has a small hourly allowance per address and one daily cap shared by everyone; when it runs out, the answer says so and links to a free account.
  • For every tool, full pages and higher limits, create a key and connect POST /mcp at https://api.prospectapis.com/mcp instead.

An API reference page for every endpoint, and docs search

  • The new API reference at /docs/reference has one page per endpoint, generated from the OpenAPI spec: parameters, request body, response codes and fields, the price, and curl, JavaScript and Python samples.
  • A search box in the docs finds any docs page, section or endpoint by a word or a path, such as webhook or /v1/funding.
  • Authentication and errors now have their own pages, with the Bearer header, the key format, every status code and the stable error codes.

Connect the hosted MCP server with OAuth

  • Claude.ai, ChatGPT and other clients that support remote MCP servers can add https://api.prospectapis.com/mcp, sign in to ProspectAPIs and approve the app, with no API key to paste.
  • A connected app is billed like any key, at the same prices, and appears in the dashboard key list as a connected app that you can revoke to disconnect it.
  • MCP clients discover the sign-in through GET /.well-known/oauth-protected-resource and GET /.well-known/oauth-protected-resource/mcp (RFC 9728). An API key still works at the same URL.

Readable MCP tool names and a delete warning

  • prospectapis-mcp 0.2.4 gives every tool a human title, such as Search Funding Rounds and Delete Watchlist, in the npm server and the hosted server alike.
  • watchlist_delete is now marked destructive, so an MCP client that honours that hint can ask before it removes a watchlist and its delivery history.

Hosted MCP server at POST /mcp

  • POST /mcp is a hosted MCP server: point any MCP client at https://api.prospectapis.com/mcp with your API key and it lists the same 14 tools as the npm package, with nothing to install.
  • Each tool call is billed exactly as the matching API route, on the key that made it.

Research briefs add a social section for X and Reddit

  • A research brief from POST /v1/research now has a social section. social.x reads the X account the company's own homepage links to: handle, followers, bio, posting cadence in posts per week, recent topics and notable posts.
  • social.reddit reads Reddit threads that are evidently about the company: a 90-day mention count, top threads with their subreddit and a summary, the communities they sit in, and a tone with its basis.
  • Every value cites the post or thread it came from. Counts, followers and cadence come from the platform data, not the model, and a platform with nothing found is left out. It is part of the same brief, with no new request field.

Blog, about page and agent discovery files

  • The blog is live, starting with a guide to sales prospect research.
  • llms.txt lists the home page, the blog index and every post.
  • New /about and /changelog pages.
  • Agents can discover the API from /.well-known: an API catalog, an agent card, an MCP server card, an agent skills index and a security.txt.

MCP server on npm

  • prospectapis-mcp 0.2.0 is published on npm under the MIT license. Run it with npx -y prospectapis-mcp and your API key; it exposes one tool per customer route, 14 in all.
  • New tools in 0.2.0: research_submit and research_get for prospect research, six watchlist tools, feedback_send and pricing_get.
  • The /mcp page has install steps for Claude Code, Claude Desktop, Cursor and any client that reads an mcpServers config.

Public launch: site, dashboard, prices and watchlists

  • prospectapis.com is live with a dashboard: sign up, create up to 25 API keys and buy prepaid credits from $10.
  • GET /pricing returns the live price list without a key. The site reads its prices from it, so a page can never show a price the API is not charging.
  • Prospect research is open: POST /v1/research is priced at $0.15 per completed brief, held when you submit and refunded if the brief fails.
  • Watchlists: POST /v1/watchlists saves funding filters and a webhook, and each new matching round is posted to the webhook, signed with Prospect-Signature. Manage them with GET /v1/watchlists, GET /v1/watchlists/:id and DELETE /v1/watchlists/:id, send a signed sample with POST /v1/watchlists/:id/test and read the delivery log with GET /v1/watchlists/:id/deliveries. Each delivered round costs one record, charged only on a 2xx from your webhook.

API 0.1.0: funding signals, research in early access and feedback

  • GET /v1/funding searches funding rounds by date, round, amount, company, domain, investor and country, with cursor paging. GET /v1/funding/:id reads one round. Every record carries the URL of the announcement it came from.
  • Funding records cost $0.02 each after 100 free records a month. A search that matches nothing, a 404 and a rejected filter are free.
  • POST /v1/research and GET /v1/research/:id open in early access: a cited research brief on one prospect from a domain, a website, a company name, a work email or a LinkedIn URL, in nine sections where every field lists its sources. The submit answered 503 until research was priced the next day.
  • GET /v1/account returns your credits and free records left. GET /v1/version returns the API version and GET /health whether the API is up, both without a key. Every response carries X-API-Version.
  • POST /v1/feedback sends a bug report or a feature request straight to the team, up to 10 an hour per account.

The routes named here are documented in the API reference, the tools on the MCP server page, and the prices on the pricing page. Who builds ProspectAPIs and where its data comes from is on the about page.

Written and maintained by The ProspectAPIs team

Try what shipped
with a free key.