WAKE.md for AI agents: compile project state so agents stop starting cold.
Local task-state ledger for AI coding agents. Preserve decisions, source conclusions, evidence, and checkpoints across Codex, Claude Code, Cursor, and other MCP clients.
Deepagents for Go, the easiest way to write LLM-based programs in Go
A Workspace API client library for Node.js
Automate Android APK analysis and patching with a model-agnostic, multi-agent AI pipeline using portable steering context and expert bytecode patterns.
Public-safe UI boundary for local-first agent-state observability experiments; no private runtime connectors or diagnostic claims.
v0.2.0: unpack fidelity fixes (sessionless project memories land, MCP name collisions warned, no swallowed write errors), verify/unpack structural validation gate (unsupported spec_version now rejected), tool-version single source of truth, and lockstep version bump across VERSION/--version/manifest/SPEC/site.json.
A Workspace API client library for Java
Receipt-first Rust runtime for portable agent state, trustworthy handoff, read-only re-entry, and an observable local TUI.
SKILL.state long-horizon agent-skill runtime for Hermes Agent (arXiv 2608.26263) — explicit-state execution, O(1) prompt, OpenAI-compatible backends
Git-like version control, audit trails, rollback, and branch/merge workflows for AI agent state.
A lightweight desktop companion that visualizes an MCP Agent's real-time work through character animations, status text, and completion bubbles
CNAA provides a lightweight cloud backend that allows AI agents to maintain persistent state across devices through MCP. Unlike traditional memory systems, CNAA does not perform reasoning or prompt engineering. It simply stores and serves structured agent state.
Demo of a schema-validated, Git-backed GraphQL API used as shared state for multiple AI agents
Jev-powered execution state for Pi coding agents. Track changes, checks, and blockers outside the main LLM, with evidence-backed updates and replay. Inspired by SKILL.state.
Agent State & Context Runtime Contract schemas and validation tools.
Herdr plugin that reports goose agent lifecycle state (idle / working / blocked) into your panes. goose is a line-mode CLI, so Herdr cannot detect it on its own.
Typed memory primitive for agentic PM work. The Memory rung of The PM Scaffold.
A stateful AI agent architecture using Qwen via Ollama for planning and Python for deterministic execution. Features structured JSON plans, persistent agent state, tool registry, sequential multi-step workflows, result passing between steps, and final response generation.
Agent State CI for coding agents: revert-aware invalidation and auditable context gating.
To associate your repository with the agent-state topic, visit your repo's landing page and select "manage topics."