AWS open-sources an 8-agent self-improving document processing system on Bedrock AgentCore
AWS ML Blog · rss · 2026-10-15
AWS ML Blog details an open-source agentic IDP (intelligent document processing) system built on Amazon Bedrock AgentCore, citing IDC data that document issues cost $19,732 per information worker annually and cut productivity by 21.3%.
Architecture: eight specialized agents across two graph workflows (built with Strands Agents):
- Identification: Analyzer extracts text via Textract; Matcher finds similar docs via S3 Vectors similarity.
- Training (new formats): Instructions agent generates extraction prompts → Extractor pulls structured data → Validator checks against schemas/business rules in Aurora DSQL → Instructions Fixer retries on failure; successful instructions are persisted for reuse.
- Automation (known formats): extraction with saved instructions; a Troubleshooter agent fixes data without touching instructions.
Infra: SES/S3 ingestion, EventBridge + Lambda triggering AgentCore runtime, AgentCore Gateway exposing Lambda tools via MCP, AgentCore Identity + Cognito credentials, and Observability monitoring. Deployable in 20 minutes via CDK; new document types need no manual templates as agents self-train and reuse knowledge. Full deployment steps included.
More from coding & agent
- Open-Agent-Email: free open-source alternative to AgentMail for AI agents — matchaman11 · 2026-10-15
- AWS Gen AI Weekly Redesign: One-Page Exec Summary, AgentCore Memory Goes Usage-Based — DigitalColmer · 2026-10-15
- Nous Research Launches 50+ Connectors for Hermes Agent, No Account Required — NousResearch · 2026-10-15
- Dev group chat: skip runtime LLM web search, index products via API instead — vxnuaj · 2026-10-15
- Tev-Gym: 1,170 dense-reward RL environments push long-horizon agentic training — RulinShao · 2026-10-15
- Microsoft's AutoEnvScaling lets agents build their own RL environments, lifting Qwen3 Terminal-Bench from 41.1 to 53.3 — omarsar0 · 2026-10-15