Discover and explore top open-source AI tools and projects—updated daily.
RhysSullivanLocal execution environment for AI agents
Top 64.2% on SourcePulse
AI agents often struggle with securely and reliably interacting with external systems. Executor addresses this by providing a local-first execution environment that offers AI agents a structured TypeScript runtime, a discoverable tool catalog, and a unified way to connect to various external systems like MCP servers, OpenAPI APIs, and GraphQL APIs. This allows agents to call typed functions instead of relying on insecure shell access or cumbersome API definitions, enhancing both security and developer productivity.
How It Works
Executor operates as a local control plane for agent tool usage. A local daemon indexes connected sources (e.g., MCP servers, OpenAPI documents, GraphQL endpoints) into a unified workspace tool catalog. Agents execute TypeScript code against this catalog via a CLI or an MCP bridge. A key feature is its human-in-the-loop capability: when a tool requires credentials or user input, execution pauses, initiates a secure local flow (such as OAuth or secret capture), and then resumes, keeping sensitive information out of the agent's direct context.
Quick Start & Requirements
npm install -g executor. For local development: bun install followed by bun dev.executor up. Local development: bun run executor call ....Highlighted Details
Maintenance & Community
apps/ and packages/ directories.Licensing & Compatibility
Limitations & Caveats
1 day ago
Inactive
NVIDIA
transitive-bullshit
letta-ai
Significant-Gravitas