Every row below is an autonomous agent that runs on cycle without an operator opening a screen. Click any endpoint to verify with curl. This page exists to answer one question directly: "is ATLAS just a GUI, or is it agentic?" The answer is 20 production agents, 20 of them autonomous on cycle.
| Agent · Cycle | What it does without operator pull | Endpoints |
|---|---|---|
| SAM.gov Opportunity Watcher daily |
Polls SAM.gov for federal solicitations matching ISS LLC NAICS codes and ATLAS keywords. Auto-drafts SBIR / SAM.gov response packets with pre-filled CAGE / UEI / past-performance. | /api/procurement/opportunities/api/sbir/proposals |
| Agent · Cycle | What it does without operator pull | Endpoints |
|---|---|---|
| Kill-Chain Bridge on intel update |
Translates intel-synthesis findings into Cursor-on-Target (CoT) tasks consumable by TAK / ATAK / WinTAK / JBC-P. Multiple parallel kill-chain orchestrators supported. | /api/intel/killchain-bridge/api/cot/dispatch |
| Weapons Engagement Zone (WEZ) Engine on track update |
Computes engagement zones for each interceptor type against each detected adversary platform in real time. Output feeds the operator console and the Monte Carlo planner. | /api/wez/zones/api/wez/recommend |
| Swarm Failsafe Orchestrator real time |
Detects loss of GPS, RF, or comms on any tracked friendly asset and issues failsafe procedure (return-to-base, hold, hand-off) per platform doctrine. DDIL-resilient. | /api/swarm/failsafe/status |
| Agent · Cycle | What it does without operator pull | Endpoints |
|---|---|---|
| TAK Bridge real time |
Bidirectional CoT bridge to TAK / ATAK / WinTAK / iTAK / JBC-P. Subscribes to ATLAS event bus, publishes upstream COP changes back to the warfighter end devices. | /api/tak/status/api/cot/dispatch |
| Voice Tactical Translator real time |
Real-time bidirectional translation of tactical voice traffic across allied languages (EN, UK, PL, RO, KO, JA, plus Levantine Arabic). Whisper-class STT + ATLAS LLM normalization to standard radio brevity. | /api/voice/tactical |
| Agent · Cycle | What it does without operator pull | Endpoints |
|---|---|---|
| Intel Synthesis Cycle every 6 hours |
Auto-collects open-source intelligence (Telegram, Mastodon, Bluesky, YouTube, RU/CN media, GDELT, procurement, cyber CVEs, NHC/NWS/FIRMS), runs LLM-grounded synthesis with citation enforcement, and writes a versioned synthesis report plus a threat-evolution ledger row. | /api/intel/synthesis/latest/api/intel/synthesis/history |
| DIME / PMESII Briefing Emitter on demand + scheduled digest |
Generates DIME (Diplomatic / Information / Military / Economic) and PMESII-PT framed briefings on operator-selected topics with deep-link source provenance and monthly spend cap enforcement. | /api/intel/briefing/api/intel/floating/realtime |
| Tag-Change Watcher every synthesis cycle |
Diffs successive synthesis reports per tag (system, AOR, threat family). Emits an alert the moment a confidence level, status, or countermeasure changes for any tracked entity. | /api/intel/tag-watcher/alerts |
| Source-Provenance Classifier on collection |
Classifies every collected source by reliability tier (A through F) and information credibility (1 through 6) per NATO STANAG 2511. Flags unattributed claims for human review before they reach a briefing. | /api/intel/source-provenance |
| Backtest Harness weekly |
Replays prior synthesis reports against ground truth that has since emerged. Produces a forecast accuracy score per analyst persona and per topic family. Closes the loop on hallucination and bias. | /api/intel/backtest |
| Commander Update Brief (CUB) Builder daily 0500Z |
Assembles a one-page Commander Update Brief from the prior 24h of synthesis, watchlist hits, and tag-change alerts. Emits as Markdown + PDF. | /api/intel/cub/today |
| RAG Self-Improvement Loop after each synthesis |
Indexes newly produced briefings back into the corpus so subsequent queries can ground on prior ATLAS findings. DDIL-safe corpus floor when no LLM provider is reachable. | /api/intel/rag/search/api/intel/rag/status |
| OSINT Collector Swarm every 15 to 60 minutes per source |
Eight independent collectors: Telegram, Mastodon, Bluesky, YouTube, RU/CN broadcast media, GDELT events, federal procurement (SAM.gov), and cyber feeds. Each emits canonical action records into a single delta store with corroboration scoring across sources. | /api/osint/health/api/osint/delta/stream/api/osint/corroboration |
| Natural-Hazard Watcher every 5 minutes |
Cross-fuses NASA FIRMS (wildfire), NWS (severe weather), NHC (cyclone), USGS (seismic) into a single hazard ledger with AOR clipping. Drives the COP hazard layer and triggers the kill-chain bridge for civil-event tasking. | /api/feeds/natural-hazard |
| SBIR Past-Award Intel every 12 hours |
Pulls trailing-24-month SBIR/STTR award universe from USAspending v2 (DoD/NASA/DHS contracts plus NIH/NSF/USDA grants), buckets into seven defense topics with defense-only filtering on inherently military lanes, and emits a separate Phase III follow-on tracker. Files land in docs/intel/auto/feeds-sbir-awards-* and feed the SBIR proposal engine via the intel-rag corpus loader (lane 3 — industry/doctrine). | /api/sbir/past-awards/topics/api/sbir/past-awards/status/api/sbir/past-awards/search/api/sbir/past-awards/run |
| Agent · Cycle | What it does without operator pull | Endpoints |
|---|---|---|
| AI Audit Engine on every AI call |
Records every LLM call with provider, model, tokens, cost, prompt hash, response hash, and grounding tag. Spend cap enforced per provider per month. Audit trail satisfies DoD AI governance requirements. | /api/ai-audit/recent/api/ai-audit/spend |
| AI Provider Orchestrator (Government-cloud failover) real time |
Single point of entry for every AI surface. Default chain: OpenAI GPT-4o (Azure Gov IL5) -> Google Gemini (Vertex / GDC) -> Grok via xAI (Oracle Gov, adapter ready) -> corpus-only RAG floor. Providers without classified-network authority are excluded from the default stack. Failover is visible in logs and surfaced to the caller. No silent fallback. | /api/ai/orchestrator/health/api/ai/orchestrator/query |
| Device Security Subagent every 60 seconds |
Per-device posture checks: OS patch level, MDM enrollment, biometric availability, jailbreak / root detection, configured key store. Emits to operator console; can quarantine non-compliant devices from the kill-chain bus. | /api/device-security/posture |
| Agent · Cycle | What it does without operator pull | Endpoints |
|---|---|---|
| Sensor Fusion Orchestrator real time, sub-second |
Fuses RF, EOIR, acoustic, and telemetry sensor lanes into a single track per platform. Triangulation with geometric DOP weighting. Failover when any single sensor lane drops. | /api/sensors/fusion/api/sensors/triangulation |
This page works best with your current GPS location.