[ BUILD · FINE-TUNE · ORCHESTRATE · DEPLOY ]
Twelve production-ready tools for building AI systems on AWS primitives — fine-tuning, multi-agent orchestration, retrieval, tool execution and deployment — with no proprietary lock-in and no platform between you and your own account.
Declare intent in plain JSON. TeamWeave resolves the right model, tools, and routing rules automatically.
Step Functions maps the task graph. Parallel branches execute concurrently; retry policies and timeouts are infrastructure concerns.
ToolWeave translates natural language into signed REST calls. Each tool invocation passes through risk scoring before commit.
mcp-observatory captures every PROPOSE/COMMIT pair. Structured logs, risk scores, and latency traces flow to CloudWatch & SIEM.
AIWeave is an ecosystem of open-source, AWS-native AI infrastructure tools built for engineers who need production-grade AI systems without proprietary lock-in.
Every library is composed on AWS primitives: Lambda, Bedrock, Step Functions, DynamoDB, EC2 Spot, API Gateway, S3, and Neptune. Each one runs in your own account, under your own IAM, and can be adopted on its own — nothing here requires the rest of the ecosystem.
This site is built with Tantu, an open-source React design system from the same workshop.