A local-first micro-agents runtime. One resident planner LLM receives intents and spawns short-lived, sandboxed micro-agents to do the work. Each agent gets exactly two tools, search and execute; everything else it needs lives on the filesystem and is reached through them. Sandboxing is a ladder (process, docker, microVM) so you pay isolation cost only when the threat model demands it. Every observable event is one line in a per-run JSONL audit log: if cat and grep cannot see it, it does not exist as state. Memory is plain files, no database, no vendor SDK, no cloud lock-in. Ollama by default, any OpenAI-compatible backend as fallback.
🏠
Working from home
Pinned Loading
-
plnt-work/plnt
plnt-work/plnt PublicOrchestration runtime for micro-agent workflows on Kubernetes
Python
-
plnt-work/maps-micro-saas
plnt-work/maps-micro-saas PublicGoogle Business SaaS built on plnt — reference consumer of the plnt runtime
TypeScript
-
plnt-work/microagents
plnt-work/microagents PublicPluggable workflow recipes for the plnt runtime — S3/OCI-hosted, versioned agent workflows
Python
-
gmail-mcp-server
gmail-mcp-server PublicA robust Model Context Protocol server for Gmail integration with intelligent authentication and comprehensive email operations
-
auto-apply
auto-apply PublicAutonomous job application engine — Playwright-based form filler with self-learning, multi-ATS support, and Claude Code integration
Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.
If the problem persists, check the GitHub status page or contact support.



