The Revenue Context Layer for GTM. Nous turns the raw sales activity scattered across your GTM tools into pre-computed context about every customer — identity resolved, claims sourced, buying committee and signals modeled — and serves it to your agents in one call, instead of leaving them to stitch six tools together and guess. The context graph is the engine underneath. Open source, with a hosted version.
- Every account in one record. Nous resolves the emails, profiles, and duplicates scattered across your tools into one person or company, so every agent works from the same complete account.
- Every fact carries its source. Each detail comes with where it came from and how fresh it is, so an agent acts on what it can trust and you can trace any answer back.
- The whole account in one call. An agent reads the full context in a single request instead of stitching six tools together and rebuilding it every time.
- Consistent across every agent. The more agents you run, the more this matters. They all read the same context, so the same account returns the same answer.
- Trained on your own data. Every reply, meeting, and closed deal feeds back into the graph, sharpening ICP fit on your own outcomes so it gets more accurate over time.
- Purpose-built for GTM. Accounts, buying committees, signals, and ICP fit are modeled out of the box, mapped to how GTM teams actually work.
Nous is built on three layers. Every signal from your tools lands as an observation, immutable evidence of something that happened or was said. Observations resolve to entities, one canonical record per person and company. From the evidence on each entity, Nous derives claims, the current belief about a fact with its confidence and freshness. Store the evidence, derive the truth. Throw the claims away and replaying the observations rebuilds them.
flowchart LR
OBS[Observations] -->|resolve identity| ENT[Entities]
ENT -->|derive| CLM[Claims]
classDef s1 fill:#ffffff,stroke:#6B7280,color:#6B7280;
classDef s2 fill:#ffffff,stroke:#7C3AED,color:#7C3AED;
classDef s3 fill:#ffffff,stroke:#F59E0B,color:#F59E0B;
class OBS s1;
class ENT s2;
class CLM s3;
flowchart TD
SRC[Your GTM tools: CRM, inbox, calendar, LinkedIn, enrichment, signals] -->|ingest| OBS[Observations: append-only evidence]
OBS -->|resolve identity| ENT[Entities: one per person and company]
ENT -->|derive| CLM[Claims: current beliefs with confidence and freshness]
ENT -->|connect| REL[Relationships: buying group, works_at]
CLM -->|score| DEC[Decision context: ICP fit, intent, next best action]
REL --> DEC
DEC -->|get_context, one call| AG[Your agent acts]
AG -->|record| OBS
Every signal becomes immutable evidence against a resolved entity. Nous derives the current beliefs, scores fit, and serves the whole account in one call. The outcome flows back as new evidence, so the account gets truer over time.
| Endpoint | What it returns |
|---|---|
get_context(focus, intent) |
the whole account context for a task, token-budgeted and agent-shaped |
get_account(id) |
the full record for one person or company, by email or id |
query(scope) |
filter activity across people and accounts |
One command — it signs you in, connects Nous to your coding agent, and hands off. The same command on macOS, Linux, and Windows (needs Node 18+):
npx @opennous/cli@latest initOn self-host, point it at your own instance by adding --url https://api.yourdomain.com.
Your agent now has get_context, get_account, and query. The examples below show the REST API and its JSON; over MCP your agent gets the same data as a token-budgeted summary.
The whole account context for a task, token-budgeted and agent-shaped. focus takes a domain, email, LinkedIn URL, or id, and every fact carries its confidence and freshness:
curl -X POST https://api.opennous.cloud/v2/context \
-H "Authorization: Bearer $NOUS_API_KEY" \
-d '{"focus":"acme.com","intent":"account_review"}'{
"entity": { "id": "ent_acme", "type": "company" },
"summary": "Acme Corp, ~500 employees. Sarah Chen promoted to VP RevOps 3mo ago, just deployed Salesforce. 12 SDR roles posted in 7 days. Open deal $45k, no economic buyer.",
"claims": [
{ "property": "signal.hiring", "value": "12 SDR roles in 7 days", "confidence": 0.95, "freshness": "fresh", "epistemic_class": "observed", "last_observed_at": "2026-06-10" },
{ "property": "signal.stack", "value": "Salesforce deployed 45d ago", "confidence": 0.88, "freshness": "aging", "epistemic_class": "observed", "last_observed_at": "2026-04-30" }
],
"stakeholders": [
{ "entity_id": "ent_sarah", "name": "Sarah Chen", "role": "VP RevOps" }
],
"timeline": [
{ "when": "2026-06-05T14:00:00Z", "type": "call", "tier": "brief", "summary": "competitor name-dropped" }
],
"predictions": [ { "kind": "icp_fit", "value": "high", "confidence": 0.82 } ],
"icp": { "score": 82 },
"meta": { "token_estimate": 1200, "claims_returned": 12, "claims_total": 47, "timeline_events": 9 }
}The full record for one person or company, by email or entity id. claims is keyed by property, each with confidence, freshness, and how many times it's been observed.
curl https://api.opennous.cloud/v2/accounts/sarah@acme.com \
-H "Authorization: Bearer $NOUS_API_KEY"{
"entity_id": "ent_sarah",
"type": "person",
"claims": {
"title": { "value": "VP RevOps", "confidence": 0.94, "freshness": "fresh", "epistemic_class": "observed", "observation_count": 3, "last_observed_at": "2026-05-30" }
},
"recent_observations": [
{ "kind": "event", "property": "interaction.call", "source": "fireflies", "observed_at": "2026-06-05" }
],
"icp": { "score": 82 }
}Filter activity across people and accounts with a structured scope. Add a question for semantic ranking, and set return: "entities" to get one row per account:
curl -X POST https://api.opennous.cloud/v2/query \
-H "Authorization: Bearer $NOUS_API_KEY" \
-d '{"scope":{"property":"interaction.email","since_days":30},"return":"entities","question":"accounts that replied positively then went quiet"}'{
"return": "entities",
"matched": 128,
"returned": 25,
"items": [
{ "entity_id": "ent_acme", "entity_name": "Acme Corp", "matches": 9,
"most_recent_at": "2026-05-28", "most_recent_type": "email_replied", "most_recent_source": "gmail" }
],
"rollups": { "by_type": {}, "by_source": {} },
"meta": { "token_estimate": 900 }
}Nous is operated by your agent, not by clicking through an app. The one command above (npx @opennous/cli@latest init) opens your browser, mints a workspace key to ~/.nous/config.json so there's no key to paste, and registers the Nous MCP server with the agent on your machine — it finds the Claude Code CLI and registers with it directly, or writes a .mcp.json any other MCP host reads.
Prefer to wire the MCP server up by hand? Add it to your host's config, then sign in once with npx @opennous/cli login (add --url https://api.yourdomain.com on self-host):
- Claude Code:
claude mcp add nous -- npx -y @opennous/mcp - Claude Desktop: Settings → Developer → Edit Config opens
claude_desktop_config.json(macOS:~/Library/Application Support/Claude/, Windows:%APPDATA%\Claude\). Add the block below and restart Claude Desktop:{ "mcpServers": { "nous": { "command": "npx", "args": ["-y", "@opennous/mcp"] } } } - Codex: add to
~/.codex/config.toml:[mcp_servers.nous] command = "npx" args = ["-y", "@opennous/mcp"]
- Cursor / any MCP host: add to
mcp.json:{ "mcpServers": { "nous": { "command": "npx", "args": ["-y", "@opennous/mcp"] } } }
Now hand your agent the setup itself. Tell it "Set me up, onboard my workspace and build my playbook," and it walks setup in order: profile, connect Gmail / LinkedIn / a note-taker, enrichment, import your CRM contacts.
Run the whole stack (API, worker, MCP server, frontend, Redis, and Caddy for automatic HTTPS) with Docker Compose on your own infrastructure. You bring a Supabase project (Postgres + auth) and an Anthropic API key.
Prerequisites
- A Linux server with Docker + Docker Compose
- A Supabase project (free tier is fine)
- An Anthropic API key
- Three DNS records (
app,api,mcp) pointing at your server
# 1. Clone
git clone https://github.com/NousC/opennous.git && cd nous
# 2. Configure
cp nous.env.example nous.env
# Fill in APP_DOMAIN / API_DOMAIN / MCP_DOMAIN, your Supabase URL + keys,
# and ANTHROPIC_API_KEY. Generate the encryption key:
openssl rand -hex 32 # paste the output into ENCRYPTION_KEY=
# SELF_HOSTED=true is already set, it runs the open primitive, unmetered.
# 3. Create the database
# Open supabase/schema.sql in your Supabase SQL editor and run it once.
# 4. Launch (Caddy provisions TLS automatically once your DNS resolves)
docker compose --env-file nous.env up -d --buildOpen https://app.yourdomain.com and create the first account, and it becomes the owner. To close public registration afterward, set DISABLE_SIGNUPS=true in nous.env and re-run ./update.sh. Update any time with ./update.sh (it pulls the latest, rebuilds, and flags new DB migrations).
Point your agent at your instance. On self-host, pass your own API URL to the same one command — it signs you in against your server, registers the MCP pointed at it, and saves the key (plus the URL) to ~/.nous/config.json, which the MCP reads automatically:
npx @opennous/cli@latest init --url https://api.yourdomain.com→ Full walkthrough in the self-host guide.
For local development against your Supabase project without Docker:
git clone https://github.com/NousC/opennous.git && cd nous
cp .env.example .env # fill in Supabase + Anthropic keys
pnpm install && pnpm dev| Layer | Stack |
|---|---|
| API | Node.js (ESM), Express |
| Frontend | Vite, React, shadcn/ui |
| Database | Supabase (PostgreSQL + pgvector) |
| MCP | @modelcontextprotocol/sdk |
| AI | Anthropic Claude |
| Package manager | pnpm workspaces |
We love contributions. See the Contributing Guide before opening a PR.
Nous is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). You are free to use, modify, and self-host it. If you run a modified version as a network service, the AGPL requires you to make your source available to that service's users. Nous Cloud runs this same open core, hosted and managed, with the team layer (CRM sync, lead lists, the ICP model) added on top. See the LICENSE file for the full text.
- We do not scrape LinkedIn or any third-party platform.
- Signal ingestion uses only official OAuth flows and approved webhooks.
- No customer data is sent to third parties without explicit configuration.