Autonomous Cognitive Agent
AEGIS
A persistent AI agent that thinks like a co-founder, not a chatbot. Multi-model routing, 3-tier memory, autonomous operations, and economic alignment -- running 24/7 on Cloudflare Workers.
Multi-Model Routing Pipeline
Every request is classified, evaluated for complexity, and routed to the optimal model based on cost and capability.
Input
Message + conversation context
Phase 1: Classify
Workers AI (free) with Groq fallback
Returns: pattern, complexity, needs_tools, confidence
Free tier
- Trust.
- Confidence ≥ 0.80. Proceed to memory lookup
- Verify.
- Confidence 0.50–0.79. Re-classify via Groq logprobs
- Escalate.
- Confidence < 0.50. Skip memory, bump executor tier
Phase 2: Procedural Memory
Lookup learned procedure by classification + complexity
Mature procedure? Use its proven executor. Otherwise fall through.
Phase 3: Complexity-Aware Default Routing
Executor Used for Cost Direct Heartbeat Free Workers AI Simple, no tools Free Groq Greetings Cheap GPT-OSS 120B Tool calling, BizOps Cheap Claude Sonnet Complex reasoning Premium Claude Opus Deep multi-step Premium Composite 4-model pipeline Multi-model
Core Capabilities
- 3-Tier Memory
- Semantic memory via Vectorize for concept retrieval. Episodic memory in D1 for conversation history. Procedural learning that improves routing over time.
- Multi-Model Routing
- Six-plus models available. Complexity-aware classification routes each request to the cheapest model that can handle it, with confidence-based escalation.
- Autonomous Operations
- Heartbeat monitoring, goal execution, memory reflection, and operator logs. AEGIS runs scheduled tasks, manages its own agenda, and flags issues proactively.
- Economic Alignment
- Operates with a 3% equity stake mindset. Every routing decision weighs cost against quality. Thinks like a co-founder managing burn rate, not a tool burning tokens.
Live Health Check
Data from the AEGIS /health endpoint.
Figure 1 as a table
| Task | Ran | Runs | Succeeded | State |
|---|---|---|---|---|
| column | about twice an hour | 48 | 48 | every run succeeded |
| dispatch | about twice an hour | 48 | 48 | every run succeeded |
| roundtable | about twice an hour | 48 | 48 | every run succeeded |
| agent-dispatch | about hourly | 24 | 24 | every run succeeded |
| agent-runner-safety-net | about hourly | 24 | 24 | every run succeeded |
| argus-analytics | about hourly | 24 | 24 | every run succeeded |
| argus-heartbeat | about hourly | 24 | 24 | every run succeeded |
| argus-notify | about hourly | 24 | 24 | every run succeeded |
| argus-routine-escalation | about hourly | 24 | 24 | every run succeeded |
| behavioral-detectors | about hourly | 24 | 24 | every run succeeded |
| ci-watcher | about hourly | 24 | 24 | every run succeeded |
| cognitive-metrics | about hourly | 25 | 25 | every run succeeded |
| consolidation | about hourly | 24 | 24 | every run succeeded |
| content-drip | about hourly | 24 | 24 | every run succeeded |
| conversation-facts | about hourly | 24 | 24 | every run succeeded |
| cost-monitor | about hourly | 24 | 24 | every run succeeded |
| cost-report | about hourly | 24 | 24 | every run succeeded |
| curiosity | about hourly | 20 | 20 | every run succeeded |
| daily-digest | about hourly | 24 | 24 | every run succeeded |
| dev-activity | about hourly | 24 | 24 | every run succeeded |
| dreaming | about hourly | 24 | 24 | every run succeeded |
| entropy | about hourly | 24 | 24 | every run succeeded |
| escalation | about hourly | 24 | 24 | every run succeeded |
| ext:self-improvement-housekeeping | about hourly | 24 | 24 | every run succeeded |
| feed-watcher | about hourly | 24 | 24 | every run succeeded |
| goals | about hourly | 20 | 20 | every run succeeded |
| heartbeat | about hourly | 24 | 24 | every run succeeded |
| inbox | about hourly | 24 | 24 | every run succeeded |
| infra-compliance | about hourly | 24 | 24 | every run succeeded |
| issue-proposer | about hourly | 24 | 24 | every run succeeded |
| nexus-contracts-warm | about hourly | 24 | 24 | every run succeeded |
| operator-log | about hourly | 24 | 24 | every run succeeded |
| pr-ready | about hourly | 24 | 24 | every run succeeded |
| product-health | about hourly | 24 | 24 | every run succeeded |
| reflection | about hourly | 24 | 24 | every run succeeded |
| routine-heartbeat | about hourly | 24 | 24 | every run succeeded |
| runtime-guard | about hourly | 24 | 24 | every run succeeded |
| self-improvement-housekeeping | about hourly | 24 | 24 | every run succeeded |
| social-engage | about hourly | 24 | 24 | every run succeeded |
| sprint-pipeline | about hourly | 24 | 24 | every run succeeded |
| tarotscript-loop | about hourly | 24 | 24 | every run succeeded |
| task-executor-do-dispatch | about hourly | 24 | 24 | every run succeeded |
| task-reconciliation | about hourly | 24 | 24 | every run succeeded |
| workers-ai-tasks | about hourly | 24 | 24 | every run succeeded |
| ambient-gmail | about every two hours | 12 | 12 | every run succeeded |
| board-sync | about every two hours | 12 | 12 | every run succeeded |
| issue-watcher | about every two hours | 12 | 12 | every run succeeded |
| ambient-calendar | about every six hours | 4 | 4 | every run succeeded |
| analysis-dispatch | about every six hours | 4 | 4 | every run succeeded |
| ext:self-improvement | about every six hours | 4 | 4 | every run succeeded |
| ledger-reconcile | about every six hours | 4 | 4 | every run succeeded |
| self-improvement | about every six hours | 4 | 4 | every run succeeded |
| agent-events-housekeeping | once or twice a day | 1 | 1 | every run succeeded |
| backup | once or twice a day | 1 | 1 | every run succeeded |
| capability-expansion | once or twice a day | 1 | 1 | every run succeeded |
| compliance-monitor | once or twice a day | 1 | 1 | every run succeeded |
| cross-domain-synthesis | once or twice a day | 1 | 1 | every run succeeded |
| digest-noise-reducer | once or twice a day | 1 | 1 | every run succeeded |
| ext:infra-compliance | once or twice a day | 1 | 1 | every run succeeded |
| finance-monitor | once or twice a day | 1 | 1 | every run succeeded |
| identity-grow | once or twice a day | 1 | 1 | every run succeeded |
| sentinel-checks | once or twice a day | 1 | 1 | every run succeeded |
| sera-calibration | once or twice a day | 1 | 1 | every run succeeded |
| shadow-read-prune | once or twice a day | 1 | 1 | every run succeeded |
| skill-evolution | once or twice a day | 1 | 1 | every run succeeded |
| task-runs-prune | once or twice a day | 1 | 1 | every run succeeded |
| trajectory-compress | once or twice a day | 1 | 1 | every run succeeded |
| watch-auto-prs | once or twice a day | 1 | 1 | every run succeeded |
| wiki-mindspring-sync | once or twice a day | 1 | 1 | every run succeeded |
| work-item-continuity | once or twice a day | 1 | 1 | every run succeeded |
- Learned
- 1
- Proven routines: at least 3 successes and a success rate of at least 70%. AEGIS reuses them directly. One that goes unused for 14 days drops back to learning.
- Learning
- 21
- Routines still being tried.
- Degraded
- 3
- Routines that failed twice in a row, or fell under 50% success over 5 or more runs. AEGIS re-plans those tasks instead of trusting the routine. One success restores it.
- Broken
- 0
- Routines taken out of use.
Tech Stack
Explore AEGIS
AEGIS is an ongoing project -- a persistent AI agent that evolves with every interaction. Check out the source code or head back to the projects page.