Company Signals

What it is

One row per signal: a single, materially true statement about the company right now — observed, reported, or inferred. Signals are not knowledge, not issues, and not strategy. They are the raw reality layer: “Marquis 3020 has 10 units on hand and a 27-day gap before the next PO arrives” or “Netherlands orders in the last 28 days are 3.35× the prior year”. Signals are produced by intelligence agents (the meeting extractor, the e-commerce sales agent) and by people. A curator agent reads them, links them to Processes, and turns the ones that need a human answer into Decisions — see Decision Signal Links for the evidence trail.

How it gets data

Three write paths:

Agents with the insert_company_signals / update_company_signals grants write rows during a run. The platform stamps agent_id and run_id server-side on every insert — an agent cannot set or override them, and they are stripped from updates.
Ingest — external systems push rows through the Ingest API:

POST /api/ingest/company_signals
Authorization: Bearer pk_...
Content-Type: application/json

People — in the app, or over MCP via data_row_create / data_row_update with the same grants.

tenant_id / workspace_id always come from the credential, never from the body.

Lifecycle

A signal is active until it is either resolved (no longer true, no successor) or superseded (a newer signal on the same subject replaced it — the successor carries supersedes_signal_id). Writers are expected to update or supersede an existing active signal on the same subject rather than insert a duplicate; there is never more than one active signal per subject.

Fields

Source

Field	Label	Type	Writable	Description
`source_kind`	Source Kind	text	create/edit	`meeting` / `commerce` / `marketing` / `vertical` / `agent` / `human` — which domain produced the signal
`source_id`	Source ID	text	create/edit	Id of the source row (e.g. the meeting note, the SKU) — hidden by default
`source_label`	Source	text	create/edit	Human-readable source (e.g. the meeting title or view name)
`source_date`	Source Date	datetime	create/edit	When the underlying observation was made — the default sort

Content

Field	Label	Type	Writable	Description
`signal_type`	Type	text	create/edit	`fact` / `goal` / `problem` / `decision` / `constraint` / `risk` / `opportunity` / `assumption` / `unknown`
`subject`	Subject	text	create/edit	Short, stable handle for what the signal is about (e.g. `Marquis 3020 — supply gap risk`). The de-duplication key
`statement`	Statement	text	create/edit	The signal itself — one or two plain sentences with the numbers that matter
`evidence_note`	Evidence Note	text	create/edit	Where the numbers come from (columns, windows, thresholds) — hidden by default
`process_id`	Process	uuid	create/edit	Late-bound link to the Process this reality belongs to. Empty until a curator or person assigns it

Assessment

Field	Label	Type	Writable	Description
`confidence`	Confidence	text	create/edit	`high` / `medium` / `low` — how sure the producer is
`needs_verification`	Needs Verification	boolean	create/edit	A person should confirm before anyone acts on it
`strategic_relevance`	Strategic Relevance	text	create/edit	`high` / `medium` / `low` — how much this should influence decisions

Lifecycle

Field	Label	Type	Writable	Description
`status`	Status	text	create/edit	`active` / `resolved` / `superseded`
`supersedes_signal_id`	Supersedes Signal	uuid	create/edit	The older signal this one replaces; the older row is set to `superseded` — hidden by default

Provenance (server-stamped)

Field	Label	Type	Writable	Description
`agent_id`	Agent	uuid	read-only	Agent that wrote the row — stamped on insert, never accepted from the payload
`run_id`	Run	uuid	read-only	Agent run that wrote the row — same rule

Like every workspace table, Company Signals also carries id, tenant_id, workspace_id, created_at, and updated_at. id is a registered (hidden) column so agents can target a row for update_company_signals and for supersedes_signal_id.

Query it

{
  "method": "tools/call",
  "params": {
    "name": "data_query",
    "arguments": {
      "view_slug": "company_signals",
      "filters": { "status": "active", "strategic_relevance": "high" },
      "select_fields": ["source_date", "signal_type", "subject", "statement", "confidence", "process_id"],
      "sort_by": "source_date",
      "limit": 25
    }
  }
}

Getting Started

Vertical Intelligence

AI Operating Guide

Resources

Sales & Commerce

Procurement

Marketing & Outreach

Governance

What it is

How it gets data

Lifecycle

Fields

Source

Content

Assessment

Lifecycle

Provenance (server-stamped)

Query it

​What it is

​How it gets data

​Lifecycle

​Fields

​Source

​Content

​Assessment

​Lifecycle

​Provenance (server-stamped)

​Query it

What it is

How it gets data

Lifecycle

Fields

Source

Content

Assessment

Lifecycle

Provenance (server-stamped)

Query it