No credit card. 500 free execution actions to start.
By signing up you agree to graph8's Terms and Privacy Policy.
Drafting, logging and chasing take hours that could go to buyers. Inbound leads wait for someone to reply.
The agents do that work and log every step. Your reps get the replies and calls that need a person, with the history attached.
The agents do the work. The interfaces are where they meet your buyers and your team. The builder is how you add your own plays. The controls set what the agents may do alone. All four work from the same data.
Named agents, each with one job, all working from the same data.
Where the agents meet buyers and your team: the AI receptionist, the phone, the Work app and chat.
How you add your own plays: workflows, the Workbench spreadsheet, and code.
The rules the agents work under: Autopilot, Approve or Off, a log of every action, and undo.
Fifteen agents, each with one job, all working from the same data. The Chief of Staff plans the work and assigns it to the others. You choose which decisions they make alone and which come to you for approval.
Keeps pipeline stages, fields and the pipeline model correct.
Prices deals, applies the discount rules and routes approvals.
Runs the outbound campaigns and moves on to the next account.
Researches each account and buyer before anyone reaches out.
Watches for intent, site visits, hiring and job changes.
Fills in missing contact and company data.
Writes the emails, messages and call scripts.
Books meetings and handles reschedules.
Reports replies, meetings and pipeline.
Fixes duplicates and incomplete records in the CRM.
Routes and qualifies every inbound lead.
Protects your sending domains so email reaches the inbox.
Plans the work and assigns it to the other agents.
Works the top of the funnel beside a human SDR.
Preps meetings and follow-ups for the account executive.
These agents do not work leads. They decide where a lead goes, what a record has to contain, how a deal is priced and which actions wait for a person. Every action they take is logged. Each card shows an example task.
Keeps the rules the other agents work by. It decides which rep a lead belongs to, and what a record has to contain before another agent uses it.
Prices the deal and routes it for sign-off. It builds the quote, checks it against your discount rules, and moves the stage once the terms are approved.
Reads the pipeline for the sales leader. It flags any deal that has not moved in 14 days and drafts the Monday review before the meeting.
Shows every agent action in one place. Anything you set to Approve waits here for a person to sign it off.
Six agents carry the volume. They work from the same data, so the Deal Agent reads what the SDR Agent wrote without anyone copying it across. Each card shows an example task.
Builds the list from your ICP, writes the first touch in your voice and sends it. When someone replies, the reply goes to the rep who owns the account.
Learns one rep from the mailbox, LinkedIn, calls and calendar they connect. It drafts in their tone and leaves voicemails in their voice, then hands back when a buyer wants a person.
Prepares the paperwork on an open deal: the proposal, comparable pricing, and responses to the redlines that come back from legal.
Watches your customer accounts. It reads usage and support history, flags churn risk before the renewal, and drafts the renewal brief.
Writes what the campaigns need: articles, pages, ads and emails. It works from your brand context and product proof, so the copy matches what you sell.
Drafts pages for search, then tracks two things: where they rank, and whether AI answers cite them.
Voice AI answers inbound calls and makes outbound ones. On the call it checks whether the account fits, and if it does, it books a meeting. The transcript goes on the account record, so your rep reads it before the meeting.
"We are evaluating a few vendors this quarter."
"Got it. Are you replacing a tool, or adding one?"
"Replacing two. The budget is approved."
Meeting set Thu 2pm with Alex. Transcript sent to Alex.
Each rep can have a twin. It learns from what the rep connects: mailbox, LinkedIn, calls and calendar. It drafts in their tone, leaves voicemails in their voice, and hands the live moment back when a buyer wants a person. The rep approves the twin before it runs, and can regenerate or delete it. Nothing is trained on data they did not connect.
Voicemail drop . Alex's cloned voice
Email draft
Written in Alex's tone. Waiting for Alex to approve.
Hands back
Covers Alex while Alex is busy, then hands the live moment back to Alex.
The AI receptionist answers chat, web forms, email and phone. It replies, checks whether the lead fits, and books a meeting when it does. Every exchange is logged on the account and goes to the rep who owns it.
"Do you integrate with HubSpot?"
Demo form . enterprise tier
"Send pricing for 50 seats"
Inbound call . fits the ICP
Open the graph8 Work app on your phone and message your agents and your twin there. Or message graph8 from Roam, Slack, WhatsApp or iMessage. The agents make the change in graph8, and every action is written to the account record and the log.
Add northwind to the Q2 expansion launch and draft the first email.
Done. Added northwind (94 ICP fit). Drafting now.
Executed in the system
Account added . first-touch email drafted . queued for your approval.
Start a workflow from a buyer signal or a schedule. Lay out the steps once, and an agent runs them in order. Add an approval at any step where you want a person to sign off. Reuse the same workflow across campaigns.
Account hits in-market score 85+
Enrich contacts and buying committee
Draft first-touch in brand voice
SDR Agent sends and routes replies
Workbench looks like a spreadsheet. Add a column, write what it should contain, and the agents fill it in for every row: an account score, a drafted line, or the next step.
The agents and data the app runs on are also available over an MCP server, a REST API and a CLI. Connect Claude, Cursor or ChatGPT to your graph8 MCP server and they get first-party tools that can read and write, not a read-only connector. This is programmable revenue: the platform, plus the APIs, MCP server and SDKs to run it and build on it.
$ claude mcp add graph8
connected . 100+ first-party tools over OAuth
$ g8 accounts find --icp 90+ --in-market
46 accounts matched . list built
$ g8 launch start q2-expansion --approve-first
sequence queued . waiting for your sign-off
The agents change real records. They move deals, send quotes and update fields. Every action is logged, so you can see what happened and undo it.
Every agent works from the same data, so the Deal Agent reads the SDR Agent's notes. Nothing is copied between tools.
They make the change in graph8: move a deal, send a quote, update a record. They do not stop at a suggestion.
Every action is written to a log that is never edited, so it can be replayed or undone. You see who did what and when.
Pick one for each decision type. Autopilot runs it, Approve waits for a person, and Off keeps it manual.
The agents carry the volume. When a buyer wants a person, your rep gets the conversation with the buyer's history and a drafted next step.
Built on a decade of CIENCE campaign data, across thousands of B2B campaigns.
Each one has its own page with what it does, what your team still does, and an example.
The agents find demand, send, follow up and qualify. When a buyer wants a person, the call or meeting goes to your rep with a brief: the buyer history, every signal and a suggested next step. You choose which decisions the agents make alone.
The agents do the volume. Research, first touches, follow-ups and replies, across every channel.
Your rep takes the conversation. The call, the meeting and the close stay with a person.
Every handoff comes with a brief. The buyer history, the signals and a suggested next step are on screen when the rep picks up.
"How does this compare to what we run today?"
They run six tools today. Ask which ones they would drop first.
One buyer signal, followed through the five steps every agent on this page uses.
A buyer visits /pricing three times. The agent sees it.
It reads the account history and the log to pick the next step.
It drafts the email, books the call or updates the record. If the step is set to Approve, it waits for you.
It writes the outcome to the log, where you can read it and undo it.
The result shapes the next play, for this agent and the others.
The SDR and Deal agents build territory lists, draft first touches in your voice, prep each call with the buyer's history, and keep the CRM up to date. Your reps take the calls and approve what goes out.
Reps spend more of the day with buyers and less on admin.
The agents are built on ten years of CIENCE campaign data, across thousands of B2B campaigns. They find buyers in 700M+ contacts and 100M+ companies, with intent and visitor signals.
Execution credits send, call, and run agents. Contact credits reveal and enrich the people you target. Three ways in.
A free entry point to the full platform. No credit card.
Unlimited contact data, in the tools you already use.
Unlimited users and data. Every app, one bill.
The $99 Team plan is the full platform: everything below, for the whole org. MCP Unlimited and Pay as you go are lighter entry points with usage and seat limits.
Execution credits send, call, and run agents. Contact credits reveal and enrich the people you target.
Sign-up takes about ten seconds with Google or Microsoft. Then an agent sets up the account with you in a conversation: your calendar, mailbox, CRM or phone number, and your website. No card and no demo.
One email. Two weekly sends: the Tenbound research report with the invite to the Tuesday community call, and the graph8 changelog with the invite to the weekly demo. Written by the Institute and the builders. Unsubscribe any time.