> ## Documentation Index > Fetch the complete documentation index at: https://docs.flowla.com/llms.txt > Use this file to discover all available pages before exploring further. # REX > Read the deal intelligence REX builds — signals, scores, insight cards, and the evidence behind them — and write signals back. [REX](/rex/overview) builds a continuously updated understanding of every deal from your calls, your CRM, and room activity. Over MCP, that understanding becomes queryable: Claude can pull a deal's signals, scores, and the specific evidence behind them, and write new signals back. Everything REX exposes here reads from the room's [Knowledge Graph](/rex/knowledge-graph). Nothing is a separate or simplified copy — a score you fetch over MCP is the score on the room's Overview tab. All REX data is **read-only** except signals. `upsert_signals` is the only tool that writes. Scores, insight cards, contacts, and findings are computed by REX and can't be set from outside. ## What you can read The worklist for a deal — dated, actionable observations, each with an urgency level and recommended next steps. One call for the whole picture: Deal Score, MEDDPICC, insight cards, and the quote-backed findings behind them. Everyone on the room, including people added but never engaged. *** ## Signals A [signal](/rex/signals) is one specific observation REX has made about a deal, with an urgency level, a justification, and one or more recommended actions. Signals are the most useful REX data to pull over MCP, because they're already scoped to "things worth doing something about." ### Query and triage ```text One room theme={null} What signals are open on the Acme Corp room? ``` ```text Across the pipeline theme={null} Show me every high-urgency signal across all my rooms, newest first. ``` ```text Triage counts theme={null} How many open signals do I have by urgency across all rooms? ``` ```text Keyword theme={null} Find any open signals mentioning security review. ``` Leaving the room out is deliberate: `query_signals` and `aggregate_signals` both work across every room you have access to when no `roomId` is given, which is what makes a morning pipeline sweep a single call. Ask for counts before details. `aggregate_signals` returns just the urgency breakdown (for example *4 low, 2 medium, 1 high*), so Claude can tell you where to look without pulling every signal body first. ### Signal properties | Property | Values | | ------------------- | ------------------------------------------------------------------------------ | | Urgency | `HIGH` · `MEDIUM` · `LOW` | | Status | `TODO` · `IN_PROGRESS` · `DONE` · `DISMISSED` | | Justification | A capture (what triggered it) and an importance (why it matters for this deal) | | Recommended actions | One or more `{ action, actionDescription }` next steps | ### Writing signals `upsert_signals` lets an external agent or workflow add its own signals to a room's feed, or update ones it created earlier. A signal you write appears in the room's feed and in the org-wide **REX's signals** view exactly like a REX-generated one. ```text Create theme={null} Add a high-urgency signal to the Acme room: the renewal date moved up to September, so the security review needs to start this week. ``` ```text Update theme={null} Update that signal to medium urgency — the review is now scheduled. ``` Updating requires the signal `id` you got back when you created it. There's no dedupe on title, so calling `upsert_signals` twice without an `id` creates two signals rather than replacing the first. *** ## Knowledge graph [`get_knowledge_graph`](/mcp/tools#rex-knowledge-graph) returns a room's whole picture in one call: its [Deal Score](/rex/deal-score), the eight [MEDDPICC](/rex/meddpicc) dimensions, the four [insight cards](/rex/insight-cards), and the quote-backed **findings** behind all of them. There's no separate call per view, so a question that spans several of them costs the same as a question about one. ```text Deal Score theme={null} What's the deal score on the Globex room, and what's dragging it down? ``` ```text Full MEDDPICC theme={null} Give me the MEDDPICC breakdown for the Acme room. ``` ```text One dimension theme={null} Why is Paper Process scored so low on the Acme deal? ``` ```text Narrative read theme={null} Summarise where the Acme deal stands using REX's insight cards. ``` ```text Across rooms theme={null} Compare the MEDDPICC Champion scores on my three biggest open deals. ``` The eight MEDDPICC dimensions are `metrics`, `economic_buyer`, `decision_criteria`, `decision_process`, `paper_process`, `identify_pain`, `champion`, and `competition`. The four insight card categories are: | Category | Covers | | ------------ | ------------------------------------------------------------------------ | | `motivation` | What's driving the deal — pain, problem, time pressure | | `stance` | Where you stand — competitive position, perceived value | | `obstacle` | What's putting the deal at risk — concerns, blockers, unmet requirements | | `trajectory` | Who decides and what's next — the buying path and committed next steps | A room without enough activity yet returns `notScored` rather than a `0`. Those are different things — `notScored` means REX doesn't have the evidence to judge, while `0` is a judgment. Don't let a summary flatten one into the other. ### Findings Findings are the quote-backed moments REX has picked up from meetings — the raw material every score and card is built from. Each one carries a type, a description, the exact quote, and the meeting and contact it came from, which is what lets you answer questions no summary view covers. ```text By type theme={null} List every unanswered question in the Acme room, with who asked it. ``` ```text By person theme={null} What objections has Sarah raised across all our calls with Globex? ``` ```text Commitments theme={null} What has the Acme buyer promised that they haven't delivered yet? ``` ```text Competitive theme={null} Every time RoutePath came up on the Globex deal, and what was said. ``` The knowledge graph and signals answer different questions. The graph gives you the narrative and the evidence for a deal review; signals give you the worklist. Pull the graph when you want to understand a deal, signals when you want to act on it. *** ## Room contacts [`get_contacts`](/mcp/tools#contacts) returns everyone associated with a room (pass `roomIds`), enriched with view counts and company associations — not just people who've shown up. Omit `roomIds` to get every contact in the organization instead of one room. ```text Everyone theme={null} Who's on the Acme room, and which of them have actually opened it? ``` ```text Never engaged theme={null} Which contacts on the Globex room have never visited? ``` *** ## Putting it together REX tools compose well with the room tools. A few things worth asking for in one go: ```text Pre-call prep theme={null} Before my Acme call: give me the open signals, the insight cards, and anything the buyer committed to that's still outstanding. ``` ```text Pipeline sweep theme={null} Across all my rooms, show me high-urgency signals grouped by room, and flag any deal where the score is below 3. ``` ```text Close the loop theme={null} For the Acme room, list the unanswered questions REX found and add an action item for each one so they don't get lost. ``` ```text Stakeholder check theme={null} Who's on the Globex room but has never opened it, and has anyone on a call mentioned them? ``` REX's scores and statuses are AI assessments graded from evidence, not facts. Hard data like a close date or a CRM field value is captured as-is, but severity, urgency, and sentiment are judgments. Anything you pull over MCP carries its justification and source for exactly that reason — check the evidence before acting on the number. ## Full parameter reference See the [REX signals](/mcp/tools#rex-signals), [Contacts](/mcp/tools#contacts), and [REX knowledge graph](/mcp/tools#rex-knowledge-graph) sections of the [tool reference](/mcp/tools).