The AI lead research agent that knows when to reach out

Leadex is an AI-powered B2B lead research tool that helps SDRs, RevOps teams, and founders build, enrich, and import contact lists using a chat-native agent that plans and executes research across the open web.

You describe a target audience in plain English - for example, "Series A fintechs in France with 20+ engineers, pull VP Engineering and Head of Growth." Leadex returns a step-by-step research plan. After approval, it searches company and people databases, opens websites with a cloud browser agent, enriches contacts through Apollo and the other connected enrichment providers, and optionally pushes results into whichever CRM you've connected.

Every completed job produces one deduped CSV, a numeric summary (rows, sources visited, duration), and a transparent live log. The roster of supported CRM and enrichment integrations is growing every week. Leadex is developed in Lisbon, Portugal and operates at getleadex.com.

Leadex turns a plain-English audience brief into a researched lead list in four stages: describe, plan, execute, deliver.

First, you describe your ideal customer profile in one chat message. A planner LLM then drafts a step-by-step plan naming concrete sources (Crunchbase, company websites, LinkedIn) and tools (web search, browsing, enrichment, CRM push). You approve the whole plan or send a new message to re-plan - there is no per-step editing.

On approval, Leadex executes steps sequentially while streaming a live log of URLs visited and rows extracted. Intermediary CSVs drop after each step, so partial results are never lost. The final output is one deduped CSV plus a short summary, and a Stop button is available at any point during execution.

Yes - signals are a first-class filter, trigger, and enrichment column inside a Leadex research plan. Supported signal types include funding rounds, hiring spikes, job changes, leadership promotions, product launches, tool and tech-stack adoption, and event sponsorship.

Ask for "Series A fintechs that hired a VP Engineering in the last 90 days" or "SaaS companies whose CTO just changed jobs and run Shopify" and Leadex reads news, hiring boards, LinkedIn, and Crunchbase to build the list.

Unlike intent-panel vendors that sell pre-computed buyer-in-market scoring, Leadex composes signals from public sources on demand - no pre-built panel, no per-signal module fee, source URL on every row.

Leadex is an alternative to Apollo.io for teams that want open-web discovery and chat-native prospecting rather than a filter-based database UI.

Apollo.io is a contact database with filter dropdowns (industry, title, headcount, location) over its own B2B data graph. Leadex is a research agent: you describe the audience in plain language, and it searches the public web, opens company websites and LinkedIn profiles, and extracts structured rows. Leadex does not host its own contact database. Instead, it uses your Apollo API key - along with whichever other enrichment providers you've connected - to enrich emails and phone numbers on rows it has already discovered.

Teams often keep Apollo for bulk contact export and add Leadex for ICP-specific lists that filter dropdowns cannot express.

Full comparison: Leadex vs Apollo.io ->

Leadex and Clay both automate lead research with AI, but they differ in interface and workflow.

Clay is a spreadsheet-first enrichment platform: you start with a table, add columns backed by data providers or AI prompts, and fan out enrichment per row. It excels at deterministic, column-by-column workflows. Leadex is chat-first: you describe the audience once, approve a research plan, and get back a deduped CSV plus a live execution log. There are no tables to configure and no columns to wire.

Clay suits RevOps teams that want to compose reusable enrichment waterfalls. Leadex suits SDRs and founders who want a single prompt to produce an ICP list without building a spreadsheet first.

Full comparison: Leadex vs Clay ->

ZoomInfo is a closed B2B data vendor selling seat-based access to a proprietary contact and company database with filter and signal-based search. Leadex is an AI research agent that works across the open web and lets you bring your own enrichment provider.

ZoomInfo owns the data; you rent query access. Leadex owns no contact database. It composes semantic web search, a cloud browser agent, and your connected enrichment providers (Apollo and others) into a per-run research plan, so your lead sources are whatever sites actually host the information - Crunchbase, news, conference pages, LinkedIn, company sites.

ZoomInfo is often chosen for scale and compliance-vetted contacts. Leadex is chosen for ICP flexibility and vendor-neutral pricing.

Full comparison: Leadex vs ZoomInfo ->

Leadex works as either, depending on your existing stack. As an add-on: Leadex supplements Apollo by discovering companies and people Apollo's filters cannot express - for example, "SaaS companies that sponsored the last three RSA conferences." You connect your Apollo API key under Settings -> Credentials, and Leadex calls Apollo bulk people enrichment and organization enrichment on the rows it has discovered.

As a replacement: teams without an Apollo subscription use Leadex to find companies via web search, extract contact hints from company sites and LinkedIn via the browser agent, and import directly into whichever CRM they've connected. The trade-off is that standalone Leadex cannot match Apollo's contact-level coverage.

Most teams run both side by side.

Common B2B lead enrichment tools include Apollo.io, Clay, ZoomInfo, Lusha, Cognism, Hunter, and Leadex. Apollo and ZoomInfo are database vendors - you query their contact graph. Clay is a spreadsheet-style waterfall builder that chains multiple enrichment providers. Lusha and Hunter focus on email and phone lookups at the individual-contact level. Cognism competes in EU-compliant B2B data.

Leadex sits adjacent to this category: it is a research agent that discovers leads on the open web and then routes contact-level enrichment to Apollo and the other enrichment providers you've connected, using the API keys you supply. The list of supported providers is growing every week. Teams often pair Leadex (discovery) with Apollo or Clay (enrichment) rather than picking one tool. Each has a different sweet spot: database size, geographic coverage, enrichment accuracy, pricing model, and interface.

Tools similar to Apollo.io include ZoomInfo, Cognism, Lusha, RocketReach, Hunter, Clearbit (now HubSpot Breeze Intelligence), and - adjacent to the category - Leadex. ZoomInfo and Cognism are direct competitors: closed B2B contact databases with filter-based search and seat-based pricing. Lusha and RocketReach focus on individual contact lookups. Hunter is email-first. Clearbit specialised in firmographic enrichment before being absorbed into HubSpot.

Leadex is not a drop-in replacement - it does not host a contact database. Instead, it composes open-web research, a cloud browser agent, and your own Apollo or equivalent key into custom lead-research runs. Teams that find Apollo's filters too rigid for specific ICPs often add Leadex without removing Apollo.

The best way to find B2B leads depends on how precisely you can describe your ideal customer profile. For broad, filter-expressible ICPs ("SaaS companies, 50-200 employees, US"), a database like Apollo.io, ZoomInfo, or Cognism is fastest. For narrow, signal-based ICPs ("companies that sponsored a specific conference and are hiring engineers"), a research agent like Leadex is more accurate because it reads actual web pages rather than a frozen database snapshot.

Most effective B2B prospecting stacks combine: (1) a database for coverage, (2) a research agent like Leadex for custom lists and edge-case enrichment waterfalls, (3) a CRM like HubSpot or Salesforce as the source of truth, and (4) clear rules to dedupe across all three.

An "AI SDR" usually refers to a fully autonomous outbound agent that discovers leads, writes emails, sends them, and replies - with no human in the loop. Leadex is deliberately not an AI SDR. It is a research agent with a human-approval checkpoint: every job begins with a plan the user must explicitly approve, and execution streams a transparent log so the user can stop the job at any moment. Leadex does not send emails, write outreach, or mimic a sales rep.

This design choice exists because fully automated SDRs skip the step where someone checks their work, which is where bad lists and hallucinated contacts cause reputational damage. Leadex focuses on the research phase and hands off curated CSVs or CRM imports to the human sender.

Phantombuster and Leadex both automate lead research, but they operate at different layers. Phantombuster is a library of named "phantoms" - prebuilt scripts for specific tasks like Sales Navigator extraction, LinkedIn profile scraping, or Twitter follower lists. You pick a phantom, connect cookies, and run it on a schedule. It is LinkedIn-heavy and workflow-focused.

Leadex is a chat-native agent that composes semantic web search, a cloud browser agent, and your connected enrichment providers per run; the user never picks a script. For repeat LinkedIn-specific tasks, Phantombuster is purpose-built. For one-off custom ICPs that cross LinkedIn, news, company sites, and conference pages in a single run, Leadex is the lower-friction choice.

Full comparison: Leadex vs Phantombuster ->

See the full field guide: Leadex vs Apollo, Clay, ZoomInfo & Phantombuster ->

To find B2B leads with Leadex, send a single chat message describing the target audience - company shape, geography, signals, and the contact titles you want.

Example input: "Find Series A fintechs in France with 20+ engineers. Pull VP Engineering and Head of Growth. Push to HubSpot, tag fr-fintech-q2."

Leadex returns a plan with typical steps: (1) semantic search over Crunchbase and news for matching companies, (2) browse each company site to extract headcount and tech focus, (3) enrich via Apollo to find the named titles with email and phone, (4) push to HubSpot with the specified tag. After you click Approve, a live log streams into the chat and the final deduped CSV lands in the same thread.

Leadex can extract structured data from LinkedIn profile pages via its cloud browser agent, but email addresses are usually not present on public LinkedIn profiles. The standard flow is: Leadex opens a LinkedIn URL, extracts the name, title, company, and visible profile fields, then passes those rows to an enrichment step that calls Apollo's people-enrichment API.

Apollo returns the work email, phone, and seniority where available. You must supply your own Apollo API key. For company-wide contact discovery, Leadex can also take a domain list and run Apollo People Search with title or seniority filters - for example, "find all VPs of Engineering at these 200 companies" - without visiting LinkedIn at all.

To enrich company data from a domain using Leadex, send a message like: "Enrich these domains: stripe.com, vercel.com, linear.app. Add headcount, industry, funding, and HQ location." The planner produces a single enrichment step that calls Apollo's organization bulk-enrichment API with your company-scoped Apollo credential.

Output columns include headcount, industry, keywords, founded year, total funding, last funding round, HQ city and country, and company LinkedIn URL. Rows are returned as a CSV. You can also reference a prior CSV in chat; Leadex reuses rows from the most recent job artifact in the same thread. Within Apollo's per-batch limit, company enrichment typically completes in under 60 seconds.

The fastest way to build a lead list with AI using Leadex is to describe the list aloud, approve the plan, and download the resulting CSV. Step 1: write your ICP as one sentence - include industry, geography, headcount, signals, and contact titles. Step 2: review the plan Leadex returns. It names the sources (Crunchbase, company websites, news, LinkedIn) and tools used.

Step 3: click Approve and watch the live log. Step 4: download the CSV attached to the final chat message. A follow-up message in the same thread carries full context - "now exclude the ones already in our CRM" works because Leadex sends the entire conversation history to the planner on every message.

Leadex is best at producing one-off, high-precision lead lists whose ICP cannot be expressed in filter dropdowns. Examples it handles well: "Series A fintechs in France whose Head of Engineering posted about hiring in the last 90 days," "all companies that sponsored RSA 2025 and have more than 50 engineers," or "European agencies that publish case studies about Shopify migrations." Traditional databases cannot filter on these signals; Leadex composes semantic search, website extraction, and enrichment to find them.

It also handles mundane tasks fast: enrich a list of domains, find VP Engineering for a given company list, or scrape speaker rosters from conference pages straight into a CRM.

Yes. When you name a URL or domain directly in your prompt, Leadex opens that page with its cloud browser agent and extracts the columns you describe - no discovery step in between. Example: "Visit techcrunch.com/events/disrupt-2025 and extract every sponsor with their company name, domain, and sponsorship tier."

Leadex creates a one-step plan that opens the given URL and returns a CSV with exactly those columns. The agent can also navigate multi-page sites - pagination, tabs, category sections - and interact with pages (click, scroll, in-site search). Each task is capped at 15 minutes. For lists of URLs, Leadex chunks them into per-task batches and retries missed items once.

There is no hard-coded per-job cap on row count; the practical limit comes from the underlying tools and the 15-minute-per-task browser-agent timeout. A single Exa search returns up to 100 results per call; the planner issues multiple calls when the audience is broader. Apollo bulk enrichment is batched internally. Browser-agent loop steps chunk prior-step rows into batches and iterate, so lists of hundreds to a few thousand rows are routine.

For very large outbound campaigns (tens of thousands of contacts), run several scoped Leadex jobs partitioned by geography, industry, or title rather than a single mega-job - smaller plans are faster, cheaper, and easier to debug if a step fails.

Leadex composes four tool primitives across open-web and third-party sources. For discovery, it uses Exa - a semantic search engine indexing 50M+ company pages, 1B+ people profiles, news, research papers, personal sites, and financial reports. For page-level extraction, it uses a cloud browser agent that opens any public URL (LinkedIn, Crunchbase, conference pages, company sites, job boards).

For contact enrichment, it routes to Apollo and the other enrichment providers you've connected - bulk people and organization enrichment plus people search - using the API keys you supply. For CRM writes, it pushes into whichever CRMs you've connected through their respective APIs; the integration list grows every week. Leadex holds no proprietary contact database. All third-party credentials are stored server-side and never returned to the browser after creation.

Leadex does not run a persistent LinkedIn scraper. It uses a cloud browser agent that opens LinkedIn URLs on a per-task basis, reading only what is publicly visible to a signed-out visitor, and extracting structured fields like name, title, company, and location. There is no credential stuffing, no mass profile harvesting, and no background crawl.

Each task is scoped to a specific list of profile URLs or company pages, and sessions are torn down at job end. Rate limits are respected per provider session. For bulk contact data beyond public profile fields, Leadex routes to the enrichment providers you've connected (Apollo and others) using the API keys you supply. This is generally more reliable and more compliant than scraping for email and phone numbers.

Leadex has clear boundaries, documented so users can plan around them. It does not host its own contact database - contact-level data comes from the enrichment providers you've connected (Apollo and others) via your API keys. It does not run scheduled or recurring jobs in v1; every run is user-initiated. It does not support per-step plan editing - you either approve the full plan or send a new message to re-plan.

Browser-agent tasks are capped at 15 minutes each. The supported list of CRM destinations and enrichment providers is deliberate and growing week by week rather than exhaustive on day one. Jobs run fire-and-forget inside the Next.js server, so very large lists are chunked rather than parallelised across many workers. Any unrecoverable provider error halts the job loudly rather than silently retrying.

Leadex exposes an internal HTTP API used by its own UI, documented under /api. Endpoints include thread and message creation, job approval and stop, Server-Sent Events for live log streaming, artifact download, company and credential management, and admin read-only endpoints. Authentication uses JWT session cookies; every data route validates the session and checks company-scoped ownership.

A public partner API for third-party integration is not offered in v1 - the product is designed around the chat UI. Teams that need programmatic access typically script against the internal endpoints using a logged-in session, or fetch the final CSV artifact via the artifact download endpoint. For CRM push, the built-in CRM-import step is usually sufficient.

Yes - Leadex pushes leads directly into whichever CRMs you've connected as the final step of a research plan. To use it: store the relevant CRM credential under Settings -> Credentials, then end your prompt with "push to [your CRM]" and optionally a tag. Leadex adds a crm_import step that creates or updates contacts with the extracted columns (name, email, phone, title, company, LinkedIn URL) and returns an imported-count summary in chat.

If you prefer CSV, every completed job also produces a downloadable file in the thread. The list of supported CRMs is growing every week - check Settings -> Credentials for the current set, and write in if you need one we haven't shipped yet.

Leadex exports CSV. Every completed job produces one deduped CSV attached to the chat, plus intermediary CSVs per step so partial results survive any late-stage failure. The CSV schema is dynamic - columns are inferred from what the user asked for and what each step produced. Common columns include company name, domain, HQ location, headcount, industry, funding stage, person name, title, seniority, work email, phone, and LinkedIn URL.

Empty rows are filtered and duplicates are removed by column signature before the final write. Files are served from /api/artifacts/:key with ownership checks. XLSX, JSON, and direct Google Sheets export are not supported in v1; users convert CSV externally if needed.

Leadex uses a "loud errors, no silent corruption" policy. Any unrecoverable provider or LLM error halts the job and surfaces a clear, vendor-neutral error message in the chat. Recoverable infrastructure errors - a transient HTTP timeout, a browser-agent session blip - retry once transparently and do not surface to the user.

If a single plan step throws mid-job, Leadex logs it, skips that step, and continues with the remaining steps. The job only fails outright if zero rows were produced. Intermediary CSVs are written after every non-CRM step, so even if a late step fails you can still download partial results. Jobs are also stoppable at any moment via a Stop button while status is running.

Leadex uses an OpenAI-compatible LLM for planning, summarisation, and enrichment-parameter inference. The default planning model is GPT-4o; history compaction uses GPT-4o-mini. The model is configurable via LLM_MODEL and LLM_COMPACT_MODEL environment variables, so self-hosted or alternative backends (Azure OpenAI, compatible providers) can be swapped in at deploy time.

Every call to the planner sends the full serialized chat history of the current thread - Leadex does no selective context windowing in v1 - so follow-up messages always benefit from prior plans, results, and constraints. The LLM is used for reasoning and orchestration only. Actual lead discovery happens via Exa, the cloud browser agent, and Apollo; the LLM never fabricates contact data.

Leadex is designed for three overlapping roles. Heads of Sales and RevOps consolidate a four-tool list-building stack (discovery, scraper, enrichment, CRM import) into one prompt surface with shared company threads and vaulted credentials. SDRs and AEs skip filter dropdowns and scraper maintenance - they describe the list and get a CSV by the time coffee is cold.

Founders doing founder-led sales run the playbook without hiring an SDR, producing their first 500-lead list in one afternoon. Leadex is less suited to teams that need a pre-packaged intent panel (third-party buyer-in-market scoring of the kind 6sense or ZoomInfo sell), recurring automations, or extremely high-volume outbound (10k+/day per seat) where dedicated data vendors remain cheaper at scale.

Open-web signals - hiring, funding, job changes, product launches, event sponsorship - are a first-class part of Leadex. It reads them on demand from public sources instead of subscribing you to a closed intent graph.

Leadex suits early-stage startups and B2B growth agencies well because of its bring-your-own-key model and per-run usage. Startups without an existing sales stack describe their ICP once, connect Apollo and HubSpot keys, and produce a working lead list the same day - no seat minimums for CRM filters or vendor-locked databases. Agencies running prospecting for multiple clients create a separate company workspace per client, each with its own credentials and threads, so data stays isolated.

Because Leadex owns no contact database, it does not charge per-contact fees; cost scales with your own enrichment-provider consumption and your CRM plan. Shared threads across the team let one operator's prior runs inform the next.

Leadex pricing is available at getleadex.com; contact hello@getleadex.com for current plans. The product uses a bring-your-own-key model for third-party services: you supply your own enrichment and CRM credentials (Apollo, HubSpot, and whichever others you've connected), so external API consumption is billed directly to you by those providers. This model avoids per-contact markup. A team with an existing Apollo subscription does not pay a second enrichment fee through Leadex.

A founder without an Apollo subscription can use Apollo's free tier and still run Leadex research plans. Company-scoped credentials are shared among all members, so a single key per integration covers the whole team. Credentials are stored server-side and never returned to the browser after creation.

Leadex is built and operated from Lisbon, Portugal, and is designed to keep customer data inside the stack each customer configures. API credentials (Apollo, HubSpot, user-supplied scraping keys) are company-scoped, stored server-side, and never returned to the browser after creation. Every data API route validates the JWT session cookie and checks company-level ownership; unauthorised access returns 404 to avoid leaking resource existence.

Research runs are logged for the user in their own thread and are not reused for model training. Chat threads persist after user deletion so teammates and admins retain access; the deleted user's identifier is nullified rather than cascading deletes. See getleadex.com/privacy.html for the full policy.

No - Leadex v1 does not support scheduled or recurring jobs. Every research run is user-initiated from the chat interface. This is a deliberate scope decision. The "plan, approve, execute" loop assumes a human reviews each plan before execution, and automating that review would undermine the product's core trust property.

Users who want recurring runs today typically save a prompt template, re-paste it weekly, and approve the new plan. Scheduled runs, saved audience briefs, and webhook triggers are under consideration for later versions. Teams needing high-frequency automation today commonly pair Leadex for ad-hoc ICP research with a dedicated outbound automation tool for day-to-day sending cadence.

Yes. Leadex does not hard-code US-centric sources or English-only queries. The underlying semantic search engine auto-detects query language from the prompt, so French, Spanish, German, and other prompts return regionally relevant results. The cloud browser agent reads any language a page is written in and extracts structured fields. Apollo covers global contact data, with stronger coverage in North America and Europe than in APAC or LATAM.

Teams routinely use Leadex for EU-specific ICPs - "Series A fintechs in France," "German logistics companies with 100+ employees," "Portuguese agencies publishing Shopify case studies." Results depend on the public web coverage of each region, not on Leadex's configuration.