Brockley

Provides an open-source control plane for agent workflows that defines, validates, deploys and executes AI agents as typed graph workflows with autonomous loops, tool calling, memory and code execution, addressing production deployment, validation and observability for agent systems. Sells to developers, DevOps/SRE and IT teams in B2B organizations building production agent infrastructure. Positioned as an open-source infrastructure-as-code approach with Git versioning, Terraform deployment, CI validation and multi-provider LLM support, delivered as a self-hostable platform via Docker Compose and Helm with CLI, API, Terraform provider and visual editor.

Key features

• Autonomous superagent loops with planning
• Tool calling and reflection and memory
• Sandboxed Python code execution
• Typed graphs with JSON Schema
• 13 validation checks before deployment
• Infrastructure-as-code with Terraform
• Git versioning and drift detection
• CI/CD validation workflow
• Multi-provider LLM support
• OpenAI Anthropic Google OpenRouter Bedrock
• Durable async task execution
• Horizontally scalable stateless workers
• MCP tools via HTTP and stdio
• Runs as MCP server
• AI-programmable via SKILL.md
• CLI for validate deploy invoke
• REST API server
• Visual graph editor Web UI
• Real-time streaming with event types
• Step-by-step execution traces