Changelog

The leads table is now a grid you can shape. Resize and reorder columns, pin the ones you always read to the left, band rows by any field, and save the layout per view — it comes back the way you left it.

Switch the same list to a board when stages matter more than rows. Company logos, contact avatars, tags, and live CRM fields render inline, so you can qualify a list without opening a single record.

Opening the CRM now lands you on the pipeline instead of a menu. One brand picker per screen decides what you are looking at, and every rollup — deals, tasks, activity — respects that scope.

Contacts get their own tab, overdue work becomes a chip that takes you straight to the task, and mail stays out of the way until you ask for it.

The research engine stopped guessing its way around a site. Lead runs are routed to one of 90 source playbooks, the crawler follows links instead of stopping at the homepage, and page budgets scale with the number of leads you asked for.

Contact discovery is a first-class step now: ask for a role — founder, head of marketing, whoever owns the budget — and you get that person, with the provenance of every email on record. We also fixed a search parser that was quietly discarding a quarter of all results.

Hand the agent a campaign and it writes the sequence: subject lines, body copy, and follow-ups shaped by each lead’s company, role, and the research already on file.

Your brand’s ICP now sits in the composer, so you can see what a run will target before it starts — and override it for a single run when the brief is different.

Every reply from every connected mailbox lands in one inbox. Threads stay grouped, provider icons tell you which account answered, and you can respond without switching tabs.

Replies are classified as they arrive — interested, not now, out of office, unsubscribe — so real answers surface first and sequences stop automatically for anyone who responded.

Run two versions of a step and see which one earns the reply. A/B results sit next to open, click, reply, and bounce rates for every campaign, broken down by sequence step and sending account.

The rest of the reporting got the same treatment: rebuilt charts, a Gantt view for campaign timing, and an event calendar that shows what goes out this week before it does.

Workflows got a cleaner canvas, a template gallery to start from, and step cards that say what they do. Bounced leads are now excluded from campaigns automatically.

Billing moved to per-seat: choose your seat count in the billing dialog, see pooled outreach sending limits across the workspace, and get a readable warning before you reach a credit ceiling rather than after.

ManyPI keeps the record now, not just the list. Companies roll contacts up to the organisation they belong to, deals move through a pipeline you define, and tasks, notes, and activity live on the record they belong to.

An event engine runs underneath. When a deal changes stage, a lead replies, or a task falls overdue, automations fire — and every verified lead can file itself into the CRM the moment it is found.

Every address is verified before anything is sent to it: syntax, domain, MX records, and mailbox checks run inline, and risky or invalid addresses are held back instead of burned on a send.

Bounces are recorded against the contact and excluded from future campaigns, so your sending reputation does not pay for a bad import.

Brands let one workspace run several products or clients side by side — separate ICPs, email accounts, leads, and chat history, switched from the sidebar.

MCP connectors let the ManyPI agent call your tools, with OAuth for the ManyPI MCP server and per-tool permissions on every connection. API keys are fine-grained: give a key exactly the scopes it needs, nothing more. Projects became folders, which is what everyone was using them as.

Find the lead, write the email, send the sequence — without leaving ManyPI. Connect Gmail, Outlook, or any SMTP account, build multi-step sequences with delays and conditions, and let the agent personalise each send from the research behind the lead.

Sending caps per account, per-account scheduling, and automatic stops on reply keep campaigns paced like a person sends them, not a machine.

Bring the ManyPI agent to the AI you already use. Our new MCP Server connects ManyPI to Claude, ChatGPT, Cursor, and any MCP-compatible client — one config block, and every ManyPI capability becomes a tool your assistant can call mid-conversation.

Scoped API keys and per-tool permissions keep you in control. Find leads, scrape websites, and analyze data without leaving your chat.

Describe your dream customer in plain English — the agent hunts the entire web and hands back verified, enriched, ready-to-close lead profiles. Company sites, news, directories, and social profiles, not a stale database.

Every lead ships with contact details, firmographic context, and the reasoning behind the match. Export to CSV or push straight to your CRM.

The ManyPI Agent now plans and executes multi-step goals autonomously: it can chain scraping, lead search, and analysis inside a single run, and recover from failures on its own.

Depending on your plan, up to 10 agent runs can now execute in parallel — with live progress streamed into the chat.

Workflows got a ground-up rebuild: visual branching, conditional logic, retries with backoff, and native steps for lead search and data analysis alongside scraper runs.

Trigger on schedules, webhooks, or data changes — and deliver results to your database, warehouse, or any connected app.

The dashboard is now chat-first. Ask the ManyPI Agent for anything — scrape a website, find leads, analyze your data, or turn a site into an API — and watch it work in place.

Attach files, connect data sources, and pick your model and effort level right from the composer.

Upload any dataset — CSV, JSON, spreadsheets, or data straight from your scrapers — and ask questions in plain English. The agent explores your data, finds the patterns that matter, and explains every answer with the rows to back it up.

We are introducing Smart Discovery, an AI-powered feature that automatically identifies the most relevant webpages within any domain based on your request. It helps you target the right pages without manually searching or maintaining URL lists.

You can enable Smart Discovery when creating a scraper or activate it later for existing scrapers by clicking the Smart Discovery icon. Once enabled, every scraper run intelligently scans the root domain and selects the best-matching subpages that are most likely to contain the data you need.

Automate your entire data pipeline with ManyPI Workflows. Build reliable, end-to-end workflows that connect scraper runs, external sources, and downstream systems without manual intervention.

Set up triggers, apply conditional logic to filter and transform data, and deliver results directly to your database or data warehouse—ensuring your data is always fresh, structured, and ready to use.

Generate clear, interactive visualizations from your data with ManyPI Studio. Turn raw scraper outputs and external data sources into charts, tables, and structured views that make complex information easy to understand.

Explore trends, compare results, and share insights effortlessly—so your data does not just exist, it tells a story and supports confident, data-driven decisions.

Use ManyPI Insights to query your data with AI and receive clear conclusions and data-driven recommendations from recent scraper runs and external sources.

It helps you spot patterns, validate assumptions, and uncover opportunities without manual analysis—so you can make faster, better-informed decisions directly from your data.

We're excited to announce that ManyPI is now available in public beta! After months of development and testing with our private beta users, we're ready to open up to everyone.

Sign up today and start extracting data from any website with our powerful, AI-driven platform. No credit card required to get started.

Join thousands of developers, researchers, and data teams already using ManyPI to power their data workflows.

We've completely rebuilt our extraction engine from the ground up, resulting in a massive 72% reduction in extraction time across all types of websites.

This improvement means you can extract more data in less time, with better reliability and lower costs. Large-scale extraction jobs that used to take hours now complete in minutes.

The new engine also includes improved error handling, automatic retries, and better support for dynamic content.

Stealth Mode is now permanently enabled for all extractions. This advanced feature makes ManyPI's requests indistinguishable from regular browser traffic.

With Stealth Mode, you can extract data from even the most protected websites without triggering anti-bot measures. This includes rotating user agents, realistic browser fingerprints, and intelligent request timing.

All users benefit from this feature automatically—no configuration required.

We've started inviting our first cohort of private beta users to test ManyPI. This select group of developers, researchers, and data teams is helping us refine the platform before public launch.

Early feedback has been overwhelmingly positive, with users praising the intuitive interface, powerful extraction capabilities, and excellent documentation.

Thank you to all our beta testers for helping us build a better product!

The first version of the ManyPI Data Engine is complete! This powerful extraction engine is the foundation of our platform, capable of handling any website structure.

Built with modern web technologies and AI-powered parsing, the engine can automatically adapt to different website layouts and extract structured data with minimal configuration.

This milestone marks the beginning of our journey to make web data extraction accessible to everyone.