SCOUT for Federal Contracting

Use this when the user has solicitation or RFP text and needs SCOUT's govcon taxonomy — primary industry, work type, and sub-categories — without searching live opportunities. Pass title (required), description, and agency/office when available. Returns structured JSON classification suitable for pipeline tagging, CRM enrichment, or pre-search filtering. Pair with search_opportunities to find similar live federal/SLED bids: pass the returned work_type= and/or industry= (primary_industry) as hard filters. Keyword-only search already reranks on these fields; the filters consume them.

classify_opportunity

Use this for corporate family structure — ultimate parent, per-subsidiary obligation totals, and prime/sub teaming relationships. Companion to get_company_profile when the user needs a full family tree rather than the lightweight family block.

get_company_hierarchy_profile

Use this for a complete UEI-keyed contractor profile — identity, total prime obligations, agency/NAICS/PSC breakouts, fiscal-year trajectory, incumbent recompetes, and capability/teaming. Requires recipient_uei (12-char SAM UEI); do not name-match. Set rollup=family to aggregate across corporate siblings under the same ultimate parent. When the user states months_back (e.g. 24), pass that value — do not keep the 36-month default. recompete_window_months is a separate horizon (default 24) for incumbent expiries. The hosted company landscape includes capture rollups on this firm's full expiring book (pipeline_by_agency, pipeline_by_value_band, pipeline_by_expiry_band, pipeline_distributions), not just the capped recompetes[] table. Replaces stitching get_award_history + get_award_summary + find_recompete_contracts with client-side incumbent filtering.

get_company_profile

Inferred technographic signals for a target domain — primarily email/identity platform from public MX records (M365 vs Google Workspace). Set include_job_postings=true with company_name= for medium-confidence stack signals from public job posting snippets (P3). Label output as inferred installed-base signal, NOT confirmed full software stack.

get_company_tech_signals

Use this when the user asks for a competitive intelligence brief or contract overview for an agency — individual contracts plus market statistics, offers analysis, and pattern detection (sole-source alerts, dominant incumbents, high-competition outliers) in a single response. Supports naics_codes (comma-separated, 4-digit prefix match), min_value (default $2M), modified_within_months (default 15), exclude_vehicles, set_aside_type, min_offers, pop_end_after / pop_end_before (YYYY-MM-DD window), and place_of_performance (state, "City, ST", or region — where the work was performed).

get_contract_landscape

Use this when the user asks about federal contract vehicles (GWACs, IDIQs, BPAs, GSA MAS) — to identify competitors on a vehicle, validate vehicle fit for a NAICS, find ordering agencies, or size the market on a vehicle. Tracks 27 vehicles with task-order analytics including SEWP, ITES, OASIS+, Alliant 2, CIO-SP3/SP4, EIS, Polaris, 8(a) STARS III, GSA MAS, SeaPort-NxG, TRICARE, and MHS GENESIS. Pass a specific vehicle name or a sector keyword ("Fed Health", "DoD IT", "NASA GWACs", "GSA GWACs", "NITAAC") for a multi-vehicle rollup. Filter by ordering agency, awardee, or NAICS. Omit vehicle= to list all known vehicles. Pass awardee= without vehicle= to return that company's task-order footprint (not the vehicle catalogue).

get_contract_vehicle_info

Use this when the user asks for their morning brief, daily briefing, or "what's happening today" — returns the same unified daily intelligence the user sees on their PrimeRFP Scout dashboard and (when enabled) in their morning email. The brief covers six sections merged across all of the user's discovery scopes — the user does NOT pick a scope — sorted by deadline urgency, posted date, and value: 1. New Opportunities 2. Notable Awards & Recompetes 3. Latest News 4. Funding News 5. Agency News (filtered to scope agencies) 6. Marked Opportunity Updates (changes on watchlisted opportunities — amendments, deadline changes, awards, cancellations) For an agency-wide briefing (not user-personal), use get_agency_briefing instead.

get_daily_brief

Who already delivers this work, and whether they subcontract. Optional intent is delivery_partners, primes_to_sub, or incumbents_recompete. delivery_partners needs seller.capability_keywords or seller.uei; otherwise the default weights stay. The scores.total field is a demand-fit rank. It does not decide whether to make contact. Pass at least one of products (gazetteer names such as ServiceNow), category, psc_codes, or work_type. NAICS is a secondary filter only. Reseller channel firms are returned on excluded, not dropped. The scores.total field is a demand-fit rank. It is not a PWin and it is not a go/no-go. Dollar figures use order obligations and parent ceiling. estimated_value is not used. Subaward rows looked up by prime UEI are undated unless the source row has an action date. SAM email, phone, and fax are omitted (FOUO). limit is at most 25, and the ranked list is returned in this response.

find_demand_primes

Use when the user asks about pre-solicitation acquisition forecasts, agency planning dashboards, or opportunities before they appear on SAM.gov (DHS APFS, DOT, DOJ, GSA FCO, VA FCO, and Army OSBP workbooks as army_osbp_forecast). Returns planning data only — not live solicitations. For awarded/recompete data use find_recompete_contracts.

find_pipeline_forecasts

Use when the user asks about congressional hearing signals — NDAA, appropriations, budget testimony, procurement policy, or committee oversight affecting federal contracting. Returns structured intel chunks from transcribed hearings (not live SAM data). Defaults to the last 180 days; pass days=0 for all-time. Signals are narrative tags, not solicitation numbers — each row includes source_type/source_label and, when matched, linked_opportunity_uids from policy_intel_opportunity_links. Recency uses hearing_event_date (title/committee date, then published_at) — not ingest created_at. Pair with find_pipeline_forecasts or find_recompete_contracts for broader capture context.

find_policy_intel

Use this when the user asks about **federal (USASpending)** contracts coming up for re-bid — incumbent research, pipeline building, capture planning, or expiring-contract discovery at DoD, civilian agencies, etc. Returns incumbent, estimated value, period-of-performance end date, and days until expiry. **Do NOT use** for CEO/BD **priority lists**, **expiring landscapes**, **shortlists**, or **hosted reports** on a GWAC/MAC vehicle (OASIS, Alliant, Polaris, 8(a) STARS III, etc.) — use **generate_recompete_priority_list** instead. Optional vehicle filter: OASIS, OASIS_SB, OASIS+, OASIS_ALL, POLARIS, ALLIANT_2, STARS_III, SEWP, ITES, etc. Optional routing_target (`aas_defense`, `fedsim`, `oasis_plus`) is a candidate-generator preset for contracts that may fit a channel — it does not detect that a requirement is migrating from vehicle A to vehicle B. Set group_by_parent_idv=true for parent-IDV rollups when individual task orders are below the value threshold. Do **not** use this tool for state/local/education (SLED) or cooperative vehicles (NASPO, Sourcewell, county/municipal). For those, use find_sled_recompetes(vendor=..., state=..., within_months=...) — it supports incumbent vendor and state filters that this tool lacks. place_of_performance= filters where the work is performed (state, "City, ST", or region). It is not the incumbent's office and not the period-of-performance end date. Set-aside filter: pass set_aside= or set_aside_type= (SBA, 8(a), SDVOSB, WOSB, HUBZone, VOSB) or an exact FPDS code Scout emits (WOSBSS, EDWOSBSS, SBP, ISBEE, IEE, BI). Unrecognized labels fail closed. Every returned row includes type_set_aside. Optional naics_source=`order` keeps rows whose NAICS is on the order itself; `inherited` keeps task orders that inherited NAICS from a parent IDV. Set within_months to the horizon the user asked for (default 24, max 60): "next 6 months" -> 6, "next 2 years" -> 24, "next 5 years" -> 60. Set include_historical=true to include awards whose PoP already ended within the same window (+/- within_months). For value-targeted questions ("over $100M", "between $10M and $50M"), set min_value / max_value in raw dollars (e.g. 100000000 for $100M) — the filter is applied server-side before the detail-row cap so big-ticket contracts aren't dropped. Set sort_by="value" to rank returned rows by estimated value descending, "expiry" to rank by soonest PoP end, or omit for the default relevance score. Set limit= to how many detail rows to return (1 through your plan's max_results; default is the plan cap). The response always includes full-pipeline capture rollups (not just the capped detail rows): `pipeline_by_value_band` (<$1M / $1M–$10M / $10M–$100M / ≥$100M), `pipeline_by_agency`, `pipeline_by_incumbent`, `pipeline_by_expiry_band` (under 6 / 6–12 / 12–24 months, with counts and dollars), `pipeline_by_set_aside`, `pipeline_by_pricing_type` (FFP / T&M / cost-type mix when `contract_pricing_type` is on the cache), and `pipeline_distributions` (min / Q1 / median / Q3 / max plus histograms for award value, days-to-PoP-end, PoP-end dates, and PoP duration). Use those to answer market-shape questions without paging the row list. DATE FIELDS (per row): - `pop_end_current` / `days_until_pop_end_current` — soonest date a recompete must be in place (current PoP end, before any options are exercised). Use this for capture-planning urgency. - `pop_end_potential` / `expiry_date` (= same value) and `days_until_expiry` — latest the work could continue if every remaining option is exercised. `expiry_date` is preserved for backward compatibility with widgets; new consumers should prefer `pop_end_potential` to make the option-period semantic explicit. A contract whose options have not yet been exercised will show `pop_end_current` years earlier than `pop_end_potential` — both are real signals. VALUE FIELDS (per row): - `order_obligated` — max(net federal_action_obligation, 0) for this PIID. Never negative. - `net_obligated` — signed net (can be negative after deobligations). `deobligated` is true when net < 0. - `parent_ceiling` — IDIQ/parent current total when `parent_award_id` is set. - `estimated_value` — always equals `order_obligated` (see `filters_applied.value_basis`). Pipeline totals and value bands use the same field. Do not treat parent ceiling as the order's book. - `naics_source` — `inherited` (task order under a parent IDV) or `order`. For award history and spending totals, use get_award_history.

find_recompete_contracts

Use when the user asks about **state/local/education (SLED)** or **cooperative** contracts nearing expiration — NOT federal USASpending (use find_recompete_contracts for DoD/civilian agency awards only). Required routing examples: - "SLED contracts expiring in Virginia with Esri" → find_sled_recompetes(vendor="Esri", state="VA", within_months=12) - "NASPO / Sourcewell recompetes" → find_sled_recompetes(query="NASPO cooperative", within_months=24) - "Oregon GIS contracts expiring" → find_sled_recompetes(query="GIS geospatial", state="OR") Returns jurisdiction, incumbent vendor, cooperative vehicle, expiration date, and days until expiry. Rows include ingest tags is_gis / is_esri plus enhanced_classification when classified. When query mentions GIS, geospatial, Esri, or ArcGIS, results auto-filter to classified rows. Optional vendor= narrows to incumbent name (e.g. vendor="Esri"). Optional state= is a two-letter code. Optional work_type= / industry= (alias primary_industry=) / sub_categories= / custom_tags= are exact enhanced_classification filters. Distinctive keywords in query= are AND-matched; empty results include zero_reason and are never padded with nearest-expiry contracts. Also returned in search_opportunities when source_type is sled or all (sled_recompetes block). Set within_months from the user's horizon (default 24, max 60). sort_by="value" ranks by estimated_value when any matched row has a dollar amount; if none do, the response sets sort_applied=false and falls back to expiry instead of pretending the value sort ran.

find_sled_recompetes

Use when the user wants a **vendor-side** SLED prospect list — publishers, integrators, or custom-dev shops that hold cooperative/state awards — not open solicitations and not federal primes. Default vendor_role is publisher,integrator,custom_dev (resellers like Carahsoft/SHI/CDW-G are excluded). Default persona=capture_bd ranks by unique cooperative-master count and keeps global SIs (Accenture, Kyndryl, Oracle, Esri, …). Pass persona=vendor_partner when selling engineering capacity into GovTech vendors: that drops captive-offshore SIs, collapses NASPO state addenda to one master, and sorts 1–5 unique masters first (then value, then jurisdiction spread). Optional work_type filters on enhanced_classification (semantic), not title substring. Optional state=, vehicle=, min_contract_value=, expiring_within_months= (default 24). Examples: "who holds SLED contracts in Texas" → find_sled_vendors(state="TX"); "which GovTech software companies should we sell engineering to" → find_sled_vendors(persona="vendor_partner"). Zero matches stay empty — explain why and suggest loosening vendor_role / dropping work_type. Never substitute reseller names. For contract-level expirations use find_sled_recompetes.

find_sled_vendors

Use this when the user wants disclosed federal subawards — by prime PIID, subrecipient name or UEI, or as an aggregate across a class of prime contracts. Modes (highest-precedence argument wins): piid -> subawards under a specific prime; subrecipient_uei -> exact UEI match; subrecipient_name -> substring on subrecipient; prime_awardee_name -> who a named prime typically subcontracts to (corporate-wide; LOGCAP-scale rows dominate a name like KBR — not a program-office book); prime_naics_prefix + prime_keyword -> aggregate ranked subrecipients across a class of primes (example: prime_naics_prefix="5415" with prime_keyword="cybersecurity"). Once a program office / PIID is in play, use piid= only. Pair with get_award_history when the user starts with a prime contractor and wants the full picture.

find_subawards

Use this when the user wants teaming partners whose capabilities complement theirs on a bid — including 8(a), SDVOSB, WOSB, or HUBZone partners. Matches public teaming profiles on NAICS, capabilities, and certifications. opportunity_uid is optional. When supplied (from search_opportunities), the call is enriched with NAICS and title keywords from that solicitation; otherwise pass capabilities= (required keywords) with optional naics_code and certifications. For past prime-contract winners (not a teaming recommendation), use get_award_history. For disclosed subcontracts under a specific PIID, use find_subawards.

find_teaming_partners

Use this when the user wants a curated, AI-ranked opportunity discovery report across all active solicitations — federal, state/local, education, and commercial — filtered by keywords, NAICS, and/or agencies. report_type accepts "quick_insights", "detailed_analysis", or "trending_opportunities". Pass fresh=true to bypass any cached report. Reports typically complete in 1-3 minutes.

generate_discovery_report

Use this when the user wants a company entity brief — recompete cliff, sole-source share, subaward dollars, GAO sustain rate, NAICS rank, and corporate family — framed for compete, claim, or diligence. Pass entity as exactly one of {uei}, {slug}, or {company_name}. Facts are identical across lenses; only ordering and the closing CTA change. format is json (default) or html. Never authors figures: every number traces to a SCOUT tool return.

generate_entity_brief

Save a hosted OASIS expiring-task landscape on the account and return its URL. Same report as generate_recompete_priority_list with an OASIS preset (default oasis_fy26_dozen). Pass within_months, audience, and query to narrow it.

generate_oasis_expiring_landscape

Save a hosted recompete priority report on the account and return its URL. Use this when the user wants an expiring landscape, priority list, or CEO/BD brief scoped to a GWAC/MAC vehicle or filters. Examples: "8(a) STARS III expiring tasks next 18 months", "Polaris expiring landscape", "Alliant recompete priority list". Combines a curated or dynamic shortlist with optional pipeline stats, incumbent board, urgency bands, and capture narrative. Set report_style="executive_brief" for a two-page executive brief. Set include_synthesis=false to skip the narrative. audience: ceo | bd_lead (optional — shapes narrative emphasis) preset: curated shortlist id (default oasis_fy26_dozen); ignored when mode=dynamic query: user scope keywords (title + narrative emphasis; also parsed for filters) vehicle: GWAC/MAC filter (POLARIS, ALLIANT_2, STARS_III, OASIS, …) routing_target: candidate-generator preset (aas_defense, fedsim, oasis_plus) — not vehicle-migration detection cage_code: awardee CAGE (e.g. 05DZ5) solicitation_procedures: MAFO, fair opportunity, SB, etc. set_aside: 8(a), SDVOSB, WOSB, HUBZone, … mode: curated | dynamic — auto-selected when omitted include_synthesis: LLM narrative (default true; set false for data-only) include_pipeline: broader pipeline stats (default false for speed) group_by_parent_idv: also return parent-IDV rollups (useful for AAS Defense)

generate_recompete_priority_list

One-shot capture context: congressional policy intel + agency acquisition forecasts (+ optional recompete contracts). Cross-links by agency, NAICS, and incumbent PIID. Use for questions like DHS border tech, NDAA + pipeline, or recompete + policy together.

get_capture_intelligence_bundle

Use this when the user wants the full details for a single opportunity by its UID — contact information, attachments, set-aside, archive date, full description, and AI analysis that are not included in keyword search results. Call this after search_opportunities. Pass the uid= value from the search result row.

get_opportunity_detail

Full grant FOA detail: enhanced_grant enrich (eligibility, readiness, PO, terms), grantor profile, prior winners, more-like-this. Pass uid from search_grants.

get_grant_detail

Curated grant discovery / mission-eligible FOA shortlist for grant practitioners. Pass keywords, agencies, applicant_types, and optionally ein= (IRS EIN), ntee=, mission_lane= (food_ag|environment|youth_ed|community_dev|rural_health), state=, include_weak=. Requires grant_discovery (Tactics+). Separate from generate_discovery_report (contracts).

generate_grant_discovery_report

Aggregate counts for open grants (by agency, instrument, program area, closing soon).

get_grant_summary

Use this when you need the exact Scout taxonomy label before calling search_opportunities or get_opportunity_summary with work_type=, industry= (alias primary_industry=), sub_categories=, or custom_tags=. field is one of work_type, industry, primary_industry, sub_categories, custom_tags. Optional query= substring-filters the labels. Returns exact strings to pass as filters — do not guess NAICS/PSC from these labels, and do not put them in query=.

list_classification_values

Use this when the user asks what has changed on active solicitations — deadline pushes, status changes, amendments, modifications, or date extensions. A date window applies (maximum 30 days). change_type is the public enum (deadline_push, amendment_posted, notice_progression, set_aside_changed, classification_changed, attachment_added, attachment_replaced, cancellation, award). amendment is an alias of amendment_posted. Two modes: "summary" (default) returns aggregate stats over the window — totals, breakdown by change type, top modified opportunities, and daily change volume. "detail" returns individual change records with field-level diffs and a pre-formatted display string; requires opportunity_uid or change_type.

get_opportunity_changes

Use this when the user asks for counts, breakdowns, or volumes of currently open solicitations — federal (SAM.gov), state/local/education, and non-federal combined. Examples: "How many cybersecurity contracts are open?", "Breakdown of active IT solicitations at DoD", "How many open RFPs in Virginia?". For individual records, use search_opportunities. For historical award statistics, use get_award_summary. Filter by query (title/description keywords), naics_code, psc_code, agency, set_aside, or posted_within_days (default 90). Geography filters place of performance (SAM PopCity/PopState and the same fields other ingest writes). Publisher location is the fallback only when the notice has no place of performance. Not a title substring. • place_of_performance — where the work is performed. Same shapes as location. Wins over location= when both are set. • region — one of: midwest, northeast, new_england, mid_atlantic, south, southeast, southwest, west, mountain, pacific, conus (expands to a fixed US state list). • location — region name (e.g. midwest), 2-letter state (MO), full state (Missouri), "City, ST" (Kansas City, MO), or "County, ST" (Jackson County, MO). County requires state. • state / county / city — explicit API fields; state wins over region when both are set. Do not pass installation names or agency names as location — use state/region/place_of_performance only. source_type: "all" (default) | "federal" | "sled" | "non-federal". Filters market segment (source_category), not publisher country. Use "federal" only when the user says federal-only / SAM.gov. Notice-type: notice_types= and/or exclude_notice_types= (e.g. exclude_notice_types=Award). Response includes by_notice_type breakdown. Classification filters: work_type=, industry= (alias primary_industry=), sub_categories=, custom_tags=, min_classification_confidence= match Scout enhanced_classification exactly (same fields search_opportunities returns). Discover labels with list_classification_values. Not notice types; not NAICS/PSC. When a classification filter is set, by_work_type / by_industry are untruncated.

get_opportunity_summary

Use this when the user provides a federal PIID (prime award or task order number) and wants the full dossier — prime contract record plus subaward totals, optionally with detailed subaward rows. Prefer this over calling get_award_history and find_subawards separately when the user has a specific PIID in hand. Set include_subaward_rows=true to include up to subaward_limit detail rows (default 25, max 100). The payload includes display.durability — a SCOUT Durability contract read (six facts, never a score) — when the prime row is present. Additive keys: read_type, family_label, unit_label.

get_piid_dossier

Save a SLED and/or federal market landscape report on the account, or return questions still needed. Pass what you know; SCOUT returns clarifying questions or a report URL when ready. The full internal search template is assembled server-side — not returned to clients. markets: ``sled``, ``federal``, ``both``, or empty to receive a markets question. mission_scope: durable domain lock (e.g. ``cybersecurity``, ``construction``). Federal runs refuse a generic wedge and return ``status=questions`` with ``how_to_continue`` until locked. naics_codes: comma-separated 6-digit NAICS lock (e.g. ``237110,236220``). Threaded into every federal sub-tool (``find_recompete_contracts.naics_code``, ``get_contract_landscape.naics_codes``, ``get_award_summary.naics_code``, etc.). Required for construction/utility briefs so the hosted report is not keyword-only. recipient_uei: optional 12-char SAM UEI — prefer UEI-keyed company profile + awards over a keyword-only landscape. persona: ``price_to_win`` | ``capture_bd`` | ``pe_ma`` | ``gov_buyer`` | ``vendor_partner`` — drives title framing and section emphasis. scope_lock: when true, honor explicit mission_scope / naics and do not niche-collapse diversified firms (e.g. ``federal IT services`` must not become ``software development``). angle_note: optional freeform sender note (internal framing only; never surfaces inbound page visits or confidential diligence targets). session_state: the session_state object returned by an earlier call of this tool. Pass that object back to continue. It may include ``pws_naics_codes`` as a list. Leave it empty on the first call.

prepare_market_landscape_report

Product-firm demand intelligence: which agencies have funded demand for a named software product/category, who holds spend, and reseller channel signals. Use for software vendors (not services capture). Grounded in federal contract awards with named-product gazetteer extraction. Pass product= (e.g. "Microsoft 365", "ServiceNow") or vendor_product=. For a family of named OEMs pass products= as a comma list (or product= "OpenVMS, COBOL, Veeam") — Scout fans out one fingerprint per SKU and returns a single composite with federal $ and SLED estimated $ on separate planes. view_mode=composite (default when product+psc) adds PSC category envelope; view_mode=product is named-product slice only. include_category_envelope defaults to false (fastest first response). Set true for the PSC envelope, or use source_type=sled|all for the SLED dual-plane appendix. source_type=federal (USASpending obligations) | sled (state/local/cooperative award-registry expirations) | all (federal headline plus SLED appendix). Pass state= (e.g. TX) to narrow the SLED slice. SLED returns a dual view: named-product mentions plus a category-aligned enhanced_classification plane (not mixed into named-product totals). Does NOT confirm installed base — use get_company_tech_signals or score_prospect for inferred displacement signals.

get_product_demand_fingerprint

Portfolio product-firm landscape from award-description OEM mentions. Returns category mix, top funding agencies, PSC delivery-class mix, size bands, and dollars for awards whose descriptions matched the Scout product gazetteer. Optional agency= or category= (cyber, data_ai, zero_trust, …) filters the slice. Use for shelf/portfolio / “which agencies name these products” questions — NOT for a single named SKU (use get_product_demand_fingerprint). Not CLIN-level proof; not open SAM counts.

get_product_lens_landscape

Full program view for one named requirement: identity, status, incumbents, vehicles, funding, linked awards, notices, forecasts, and a SCOUT Durability contract read per linked active PIID. Use this when the user names a program or a program of record. Durability reads are not summed or scored. Vehicles are derived from parent IDVs on the linked awards. Membership from a recompete chain is a SCOUT grouping, not a successor the government wrote.

get_program_profile

Dated history of linked actions on one program: award actions, PoP ends, notices, and budget lines, each with its source id.

get_program_timeline

Use this when the user wants published GAO bid-protest decisions — individual case rows with decision URLs and (optionally) narrative previews. Pair with find_recompete_contracts and get_award_history to study sustain patterns at an agency. Filter by agency (use "DoD" to match Navy, Army, Air Force, Marine Corps), awardee, protester, naics_code, date range, or narrative_contains (substring on decision text). Date filters use decision_date (not the B-docket number series); rows with null decision_date are excluded from a bounded range unless include_undecided=true. Set include_narrative=true for a capped preview (returns narrative=null + enrichment_state when unenriched). Set enriched_only=true to keep only rows that carry narrative or an LLM summary.

list_protest_decisions

Use this when the user wants cross-case pattern analysis over GAO bid-protest decisions — agency sustain rates, outcome predictors by issue tag, FAR-citation effectiveness, repeat-protester patterns, incumbent-advantage signals, or tag co-occurrence clusters. Filter with agency (use "DoD" to match Navy, Army, Air Force, Marine Corps), protester, awardee, naics_code, or date range. Use focus= to refine to an issue area — examples: "technical_evaluation", "price_evaluation", "organizational_conflict_of_interest", "past_performance", "sole_source", "best_value", "disparate_treatment". Rates and predictor labels require n ≥ 5 (strong_predictor also needs rate ≥ 0.7). Date filters use decision_date and exclude null dates unless include_undecided=true. Incumbent signals omit null/nan awardees. For individual case detail use list_protest_decisions; for simple counts and sustain rates use get_protest_sustain_rates.

analyze_protest_patterns

Use this when the user wants aggregate sustain / deny / dismiss counts and the merit sustain rate — sustains divided by (sustains + denies) — over published GAO decisions. Uses the same filter set as list_protest_decisions. Merit rate is withheld when merit_decisions < 5. Date filters use decision_date and exclude null dates unless include_undecided=true.

get_protest_sustain_rates

Use this when the user asks which CRM or Chat integrations are connected, to land THIS conversation into HubSpot/Salesforce/Dynamics, to check the status of a company or pipeline deal already in the CRM, or to dry-run / Push an existing Scout feed. Call with no arguments to list THIS user's live connections, feeds, and session_land (which CRM, which Chat destinations will notify). Chat is Slack, Discord, or Teams — Teams is coming soon. Do not invent a destination as connected unless it appears in the response. Status lookup: pass match= (company, deal, PIID, solicitation, UEI) without rows. Optional lens=pipeline or accounts. If the result has needs_clarification=true, ask the user the ask= question — do not guess among candidates. Session land: pass rows from this session (recompetes or opportunities) plus optional report_id. Default limit=1 (the deal) unless they named a feed or said all/these. provider= only if more than one CRM is live. Dry-run first; push=true when they said update/write/land. Scheduled feed: pass feed_id from the list. Live write: push=true. HubSpot, Salesforce, and Dynamics receive the same Scout fields (PoP or close date, source report URL, dossier URL). Scout does not write Fit Score, PWin, or GO/NO-GO. A live CRM land/Push posts a line to each connected Chat destination (deal name, report URL, optional Slack @capture_lead).

push_crm_feed

Use this when the user asks why recompete numbers or dates from this app differ from another system they are comparing against — returns the internal methodology guide covering field mapping, PIID aggregation rules, upcoming-vs-historical behavior, and common variance causes. Does not query live award data. For live rows, use find_recompete_contracts or get_award_history.

get_recompete_methodology

Use this when a user's opportunity search returned zero results — including NAICS adjacency expansion — and they want the support team to investigate. Sends an internal notification with the search context (NAICS, query, adjacent codes searched, optional user note).

report_empty_search

Given a PIID, solicitation number, notice id, or budget-line id, return the program it belongs to, the link confidence, and the evidence. Returns none when no link exists.

resolve_program