APEX · The 21 specialist agents
When we say “21 specialist agents with ~165 graph-grounded tools,” this is the registry. Each agent has a documented intent, a tool count, a sample input, a sample output, and a median latency. No marketing-copy “agents.”
Competitive Intelligence · 7 agents
Agent 01
Intent: Compose, refresh, or compare a battlecard between two products, companies, or categories.
Sample input
“Refresh the Salesforce vs HubSpot battlecard with this week's releases and pricing changes.”
Sample output
Updated card · 14 citations · 4 new sources since last refresh · 2 pricing diffs flagged.
Top tools:
getVendorlistReleasesgetPricingDiffgetReviewSentimentcomposeBattlecardAgent 02
Intent: Aggregate closed-won and closed-lost opportunities, identify the competitor that took the deal, and summarize the pattern.
Sample input
“What competitors did we lose to in CRM deals last quarter?”
Sample output
12 losses · top loser: Salesforce (5), HubSpot (3), Pipedrive (2). Primary stated reason: 'native integration breadth'.
Top tools:
getOpportunitiesresolveCompetitoraggregateLossReasonsrankCompetitorsAgent 03
Intent: Surface what changed on a competitor's website, pricing page, product, or comms in a given window.
Sample input
“What changed on Klue's pricing page in the last 30 days?”
Sample output
1 net pricing change · Pro tier added at $35/seat/mo · Enterprise CTA replaced 'Contact Sales' with 'Book a Call' · screenshot diff attached.
Top tools:
getWebsiteDiffgetPricingHistorydiffPagescaptureScreenshotAgent 04
Intent: Detect hiring patterns, key-departure signals, and exec moves at competitor companies.
Sample input
“Has Crayon hired any forward-deployed engineers this quarter?”
Sample output
Yes · 4 FDE roles posted, 2 filled per LinkedIn signals. Hiring concentrated in NYC and Toronto.
Top tools:
getHiringPostsresolveRoletrackExecMovesrankHiringIntensityAgent 05
Intent: Monitor competitor pricing pages, plan structures, and per-seat economics over time.
Sample input
“How has Productboard's pricing changed in the last 6 months?”
Sample output
2 plan additions, 1 plan deprecated. Maker → Scale tier consolidation. Per-seat list price up 14% on Enterprise.
Top tools:
getPricingHistorydiffPricingPagescomputePerSeatTrendextractPlanFeaturesAgent 06
Intent: Read changelogs, release notes, and press to identify net-new features shipped by competitors.
Sample input
“What features has Gong shipped in Q1 2026?”
Sample output
11 features across 4 themes: AI Deal Score (Jan), Forecast Inspector v2 (Feb), Engagement Genome (Mar), Apple Intelligence integration (Mar).
Top tools:
getChangelogsextractFeaturestagFeatureThemesrankByImpactAgent 07
Intent: Resolve the persona a competitor is targeting from their copy, content, and pricing structure.
Sample input
“Who is Apollo's primary persona based on their site and pricing?”
Sample output
Outbound SDR at 50-500 seat B2B SaaS company. Secondary: RevOps at 500+ seat. Tertiary: VP Sales.
Top tools:
extractPersonaFromCopyresolvePricingPersonarankPersonasmatchToICPRFX & Proposals · 4 agents
Agent 08
Intent: Draft answers to RFP questions using approved-answer library + graph-grounded fresh facts.
Sample input
“Draft answer to: 'Describe your SOC 2 compliance posture and audit cadence.'”
Sample output
240-word draft · 4 citations to approved-answer library + 2 to compliance graph nodes · confidence 92%.
Top tools:
findApprovedAnswergetComplianceNodecomposeAnswerscoreConfidencelogForReviewAgent 09
Intent: Search, retrieve, version, and propose updates to the canonical approved-answer library.
Sample input
“Has anything in our approved-answer library gotten stale in the last 90 days?”
Sample output
14 answers flagged · 6 reference deprecated pricing, 4 reference superseded product names, 4 cite expired certs.
Top tools:
listLibraryEntriescheckFreshnessdiffAgainstGraphproposeUpdateAgent 10
Intent: Assemble a full proposal from question decomposition + draft answers + branding + cover letter.
Sample input
“Compose a proposal response for the State of California EDD RFP.”
Sample output
Full DOCX · 42 questions answered · 187 citations · cover letter + exec summary · ready for review.
Top tools:
decomposeQuestionsdraftAnswerscomposeCoverLetterapplyBrandingexportDOCXAgent 11
Intent: Parse an inbound RFP / RFI / security questionnaire into a structured list of answerable questions.
Sample input
“Decompose this 78-page security questionnaire PDF.”
Sample output
412 questions extracted · 387 mapped to approved-answer library · 25 routed to human review · time-to-respond estimate: 4 hours.
Top tools:
parsePDFextractQuestionscategorizeIntentmapToLibraryProduct · 3 agents
Agent 12
Intent: Reason about your roadmap commits against market signal, competitive moves, and customer feedback.
Sample input
“Which of our Q3 roadmap items overlaps with what competitors shipped this quarter?”
Sample output
3 of 12 commits at risk: Battlecard refresh cadence (Klue), AI deal score (Gong), category benchmarking (Crayon). Recommend reprioritizing.
Top tools:
getRoadmapmatchAgainstReleasesrankOverlapRisksummarizeFindingsAgent 13
Intent: Compare your product's feature matrix to competitors and identify net-new vs me-too vs missing capabilities.
Sample input
“Where is HubSpot ahead of us on RevOps capabilities?”
Sample output
5 capabilities: Sequence orchestration, Lead scoring v3, Forecast attribution, Snowflake-native sync, Custom report builder.
Top tools:
getCapabilitiescompareMatricesscoreGapcategorizeByAgent 14
Intent: Maintain the canonical capability map for your category · what features map to what capability, across vendors.
Sample input
“Add 'AI Deal Score' as a capability under Revenue Intelligence and find which vendors ship it.”
Sample output
Capability added · 7 vendors mapped (Gong, Clari, MEDDIC, BoostUp, Aviso, Salesloft, ZoomInfo). Pricing tier varies.
Top tools:
addCapabilitysearchVendorsByFeaturetagCapabilitylinkToProductGTM · 6 agents
Agent 15
Intent: Generate or refresh an account plan from CRM signal + product graph + persona model.
Sample input
“Draft an account plan for State Farm.”
Sample output
12-page plan · ICP score 87/100 · 4 buying-committee personas · current stack inferred · 3 wedge plays prioritized.
Top tools:
getCRMAccountinferStackbuildBuyingCommitteerankWedgePlayscomposePlanAgent 16
Intent: Construct or refine an Ideal Customer Profile from current closed-won data + market signal.
Sample input
“Build an ICP from our last 50 closed-won deals.”
Sample output
ICP: B2B SaaS, $50-200M ARR, 250-1000 employees, has CRM + RFP + CI tools. Match score against current pipeline: 38%.
Top tools:
getClosedWonAccountsextractCommonalitiesscoreFitToPipelinerankFeaturesAgent 17
Intent: Identify net-new buyer segments emerging in your category from signal patterns.
Sample input
“What new buyer segments are emerging in CI tools?”
Sample output
3 emerging segments: 'Embedded CI for PMM' (12 hiring posts), 'Agentic CI for sales' (8 startups), 'Open-source CI' (3 OSS projects with traction).
Top tools:
scanSignalClustersdetectEmergingTopicsrankByVelocitysummarizeSegmentsAgent 18
Intent: Aggregate, theme-extract, and sentiment-score reviews across G2, TrustRadius, GetApp, etc. for any vendor.
Sample input
“What are the top complaints about Productboard from the last 90 days?”
Sample output
3 themes: 'pricing tier confusion' (38% of negative), 'integration breakage with Jira' (24%), 'slow performance at scale' (19%). 247 reviews analyzed.
Top tools:
getReviewsextractThemesscoreSentimentrankComplaintsAgent 19
Intent: Generate a per-opportunity tip for an AE based on graph signal + competitor presence + prior win/loss patterns.
Sample input
“Tip for the State Farm opportunity?”
Sample output
Competitor present: Klue (LinkedIn signal · 3 contacts mention). Wedge: PYRAMYD ships graph-grounded battlecards Klue doesn't. Recommend leading with provenance demo.
Top tools:
getOpportunitydetectCompetitorPresencematchToWinPatterncomposeTipAgent 20
Intent: Run AI-Interview style discovery against a category of customers, extract themes, generate research artifacts.
Sample input
“Run discovery on 'CRM users who also use a separate CI tool'.”
Sample output
32 simulated customer interviews · 6 themes · 4 actionable insights · 1 hypothesis ready to validate with real customers.
Top tools:
runSyntheticInterviewextractThemesrankInsightscomposeReportPlatform · 1 agent
Agent 21
Intent: Maintain the 88 universal node types, custom field additions, permission model, and audit log on behalf of admins.
Sample input
“Add a 'ESG Score' field to the Vendor node, sourced from MSCI.”
Sample output
Schema updated · field added with type Number(0-100) · MSCI connector wired · 2,847 vendor rows back-filled · audit entry logged.
Top tools:
addFieldwireConnectorbackfillRowsvalidateSchemalogAuditEntry30 minutes. Bring a real question. We'll show the supervisor classifying intent, the specialist running, the multi-hop traversal, and the citation chain.