The Platform

Agents onthe dealgraph

The chat agent, the action agent and the artifact agents all run on one auditable record. Every answer names what it rests on, and says so when it does not know.

Connect in an afternoon, see impact in days.

The chat agent, the action agent, the artifact agents and the MCP server are four surfaces onto one record: each answers, acts, drafts or serves from the same dealgraph, and each names the facts it rests on.

  1. Chat agent

    What is holding Northwind up?

    Budget has not come up in 22 inputs.

    The economic buyer has missed the last two meetings.

    budget · inferred absentattendance · calendar

  2. Action agent

    High

    Get Marcus into the pricing meeting

    WhyA pricing meeting without the person who signs cannot close either gap.

    Resolves whenMarcus is on the 14 Oct invite, or a budget figure is observed.

  3. Artifact agents

    Proposal · Northwind Corp

    1 · Your situation, in your words

    2 · How we respond, criterion by criterion

    3 · Commercials

    12% discount asked · your limit is 10% · held for VP approval

  4. MCP

    observed.deal.champion

    "value": { "person": "Priya Nair" }

    "coverage": { "inputs": 14, "read": 14 }

    "producer": "champion-resolver@2.3"

one record

Your context

Your agents know how your company sells.

Your ICP, your products and pricing, your playbook and your policy limits are declared once and versioned. Every agent reads them, so an answer is not just specific to your deal, it is correct for your company.

Declared once, versioned

  • Playbookv4

    • Economic buyer met directly
    • Budget confirmed with the holder
    • Mutual plan agreed with dates
  • ICPv3

    • Logistics and supply chain
    • 1,000 to 5,000 employees
    • SAP or Oracle finance stack
  • Pricingv7

    • Per seat, three tiers
    • Discount limit 10%
    • Standard terms Net 30
  • Catalogv12

    • Platform tier
    • Integration pack, SAP and Snowflake
    • SSO through Okta

What each agent does with it

  • The action agent

    Fires on your exit criteria, not a generic checklist. Budget unconfirmed is a gap because your playbook says it is.

  • Artifact agents

    A proposal is held for approval because 12% is over your limit of 10%, and it says which limit and which version.

  • The chat agent

    Champion means what your playbook defines, so the answer matches how your team actually qualifies.

  • Your own agents, through MCP

    The same declared context reaches whatever you build, so your automations inherit the rules rather than restating them.

Declared context is versioned, so an artifact built in October still resolves against what your pricing said in October.

The action agent

From a playbook rule to Tuesday morning

Your playbook is declared once: stages, exit criteria, policy limits. Every day the action agent reads it against every deal, and each rep sees only what applies, with the reason attached. An action is a condition, not a to-do. It fires when the condition is true, resolves when the fact changes, and if a rep dismisses it the reason becomes a fact the engine reads next time.

Fires when a condition is true, resolves when the fact changes, never a to-do list that rots.

Declared once

Stages, exit criteria and policy limits live in your company context, versioned, and are read against every deal every day.

A condition, not a to-do

An action fires when its condition is true and resolves when the fact changes. Nothing sits in a list going stale.

The reason attached

Each rep sees only what applies to their deals, with why, how, and what resolves it.

Dismissals become facts

If a rep dismisses an action, the reason is recorded, and the engine reads it the next time the rule runs.

Grounded in the record

Every action rests on facts that carry their evidence, source, time and coverage. No generic advice.

The deal workspace

One workspace per deal.
Understand it, execute it, win it.

Every deal gets its own workspace, organized into two layers that work in concert. The insights layer is where the deal is understood. The execution layer is where the deal is advanced.

Insights layer

summary

The deal's single source of truth: stage, amount, close date, key contacts, and narrative status.

meetings

Meeting summaries, key insights, and risk signals extracted from every call. Plus one-click prep for the next meeting.

MEDDICC

Evidence-based methodology scoring across all dimensions, linked to specific call transcripts and emails.

stakeholders

The full buying committee mapped inside the workspace. Roles, influence, sentiment, and multi-threading gaps.

objections

Tailored responses from the company's playbook, the buyer's specific concerns, and the deal's value story.

Execution layer

dealmail

The deal's own inbox. AI-drafted emails grounded in full deal context, ready to review and send.

mutual work plan builder

Collaborative timelines between rep and buyer: milestones, owners, dates, and dependencies.

pitch builder

Deal-specific pitches built from MEDDICC evidence, stakeholder priorities, and the buyer's stated pain points.

proposal builder

One-click proposals grounded in real deal data, competitive positioning, and your value framework.

objection handling

Tailored responses from the company's playbook, the buyer's specific concerns, and the deal's value story.

dealdocs

Every artifact is saved, searchable, and proactively resurfaced when relevant. The deal's knowledge library.

competitive strategy builder

Deal-specific competitive strategy built from your battle cards, the buyer's evaluation criteria, and real-time market intel.

the chat agent

Ask anything of the record: what is holding a deal up, which commits are real, what changed since Friday. Answers cite their facts and say what they do not know.

The workspace that makes executing the deal the easiest thing a rep does all day.

Three roles. One record.

For reps

The action agent tells you what to do next, with the reason attached and the artifact ready to go. No more figuring out priorities or starting from blank.

The chat agent has read every email, listened to every call, and knows your methodology inside out. Ask it what is actually holding a deal up.

The record shows what changed on your deals since Friday, ranked, with the reason attached. Preparing for a call is reading, not reconstructing.

dealmail drafts the right email in seconds, grounded in what actually happened in the deal, not a template.

For sales leaders

The record shows which commits are real, checked against your own exit criteria, without a pipeline review. Evidence per deal, not a color.

The chat agent answers from the record across every deal. "Which of these commits are real?" Answered with the facts behind each one.

The action agent shows whether your methodology is being followed, not by checking CRM fields, but by what the record says happened.

The dealgraph reveals which deals are structurally at risk: single-threaded, missing key roles, or losing stakeholder sentiment.

For RevOps and GTM engineers

Company context turns your methodology into declared rules: stages, exit criteria and policy limits applied to every deal, versioned, without another training cycle.

The data workbench shows where the data is thin and who owns fixing it. Correct any fact and have the correction stick.

MEDDICC scores measure adherence from real deal evidence, not from CRM checkbox hygiene.

MCP exposes the same record to your own agents. Ask for a champion by name and get it back typed, with its evidence and its coverage.

See the record built from your own sources.

Connect one pipeline in an afternoon. Then ask it anything, and check what every answer rests on.