OpenNous

Build on the context graph

Connect your GTM stack with the REST API, MCP or the Claude Code plugin, and work with the context graph inside your agents.

Endpoint200 · 84 ms
GET /v2/accounts/meridian-health.com

{
  "icp": 82,
  "people": 6,
  "claims": 18,
  "sources": ["gmail","granola","hubspot","posthog"]
}
Account signalsMeridian Health
Budget confirmed for Q1call · 12 Sep
Champion replied3d ago
Stage moved to evaluationCRM
Usage down since the demo2w
Invoice paid1 Sep
Competitor named in #deals6d
Integrations
Gmail
Granola
Fireflies
Slack
Stripe
PostHog
and more
Two-way sync
HubSpot⇄account · stage · owner
Action layer
Follow-up draftedafter every call
CRM updatedaccount · stage · owner
Account handoffon closed won
Risk flaggedcontinuously
Meeting prepbefore the call
Weekly reportMonday 07:00

Trusted and reliable infrastructure

Resolved, sourced, dated and permissioned before a person or an agent reads it.

FIG 0.1

Identity resolution

One person and one company, however many systems they appear in. Resolved before anything is derived, so nobody works from a near-duplicate.

FIG 0.2

Account signals

Budget, authority, competitors, adoption and risk, read out of what was actually said and dated to the day it was said.

FIG 0.3

Sources on every fact

Every answer returns the call, email or record behind it. Your team checks a claim rather than believing one.

FIG 0.4

Permissions

A private thread stays private. Facts inherit the visibility of the record they came from, for people and for agents alike.

FIG 0.5

Freshness

Each fact carries the date of the record it came from, so a stale answer is visible as stale rather than quietly wrong.

FIG 0.6

Your data, your infrastructure

Open source under AGPL. Self-host the whole engine, or keep the raw records in a database you own.

API, MCP and plugin

One account record, served the same way to a script, an agent and the workspace.

Read the API docs →
FIG 1.1
POST/v2/context200 · 84 ms
{
  "entity":  { "id": "ent_meridian", "type": "company" },
  "summary": "Pilot at $40k for Q1. Priya Nair is the
              economic buyer. DPA review open with IT.",
  "claims": [
    { "property": "deal.budget", "value": "$40k for Q1",
      "confidence": 0.94, "freshness": "fresh" }
  ],
  "icp":  { "score": 82 },
  "meta": { "token_estimate": 1200 }
}

One call, the whole account

REST: summary, claims with confidence and freshness, people and timeline. focus takes a domain, an email, a LinkedIn URL or an id.

FIG 1.2
every day · 09:00Agent #deals
Claude
MCP · get_context

MCP server

Give Claude, ChatGPT, Cursor or your own agent the account, on a schedule or on demand.

FIG 1.3
query("deals that went quiet")
Northwind Labsno reply · 14d
Brightpathcompetitor named
Halden & Conew stakeholder
matched 12 · returned 3

Query every account

Ask across the whole book in one call, with a question for ranking and one row per account back.

FIG 1.4
claude code
› /opennous:brief meridian-health.com● get_account · 41 claimsBudget $40k for Q1 [call]› /opennous:reach-out priya

Claude Code plugin

Briefs, follow-ups and pipeline reviews as slash commands.

FIG 1.5
record("priya@meridian.health")
Asked for the security pack
claim · source call/2026-10-01
on the record · every agent sees it

Record what happened

An agent records what it learned against a precise person or account, with its source, for every other agent and person.

Start building on the context graph

Connect your systems, take an API key, and read your first account.