What is MyAgent?
MyAgent is a Python-based local agent runtime for people building or studying AI agents, including developers creating personal, project-analysis, or domain-specific assistants. It coordinates model requests, tool calls, permissions, memory, reusable Skills, external MCP tools, and child agents in a single execution loop. More than a chat interface, it provides a runnable command-line environment with session saving, recovery, and connections to OpenAI-compatible or Anthropic-compatible model APIs.
MyAgent Features:
Agent Execution Loop: The runtime parses model tool calls, checks permissions, executes tools, returns results for further reasoning, and saves the session.
Model Protocols: Connect to models through OpenAI-compatible or Anthropic-compatible interfaces, with configuration available through environment variables or command-line options.
Tools and Permissions: Built-in tools cover file reading and writing, precise edits, search, Shell commands, Skill calls, child agents, and MCP tools.
Reusable Skills: Load project-level and user-level SKILL.md files, retrieve relevant Skills, and run them inline or in a forked execution.
Skill Evolution: Extract reusable rules from user feedback and add or merge them into Skills, with version snapshots and audit records.
Memory and Sessions: Store long-term memory by project path, save sessions for later recovery, and compact context or truncate large tool results.
MyAgent Benefits:
Inspectable Workflow: A visible runtime loop makes it easier to understand where model reasoning ends and tool execution, permissions, and result handling begin.
Controlled Changes: Plan Mode lets users analyze and prepare a plan before approval, while blocking write operations and Shell commands during planning.
Reusable Know-How: Stable guidance can carry forward into similar tasks, reducing the need to restate the same working preferences each time.
Project Continuity: Project-specific memory and resumable sessions help preserve useful context between tasks instead of requiring users to start over.
Flexible Foundation: A local runtime with extension points gives developers a base they can adapt for personal assistants or specialized project workflows.
Use Cases:
Project Analysis: Ask an agent to inspect a codebase, explain how its main modules fit together, or trace how its agent loop works.
Multi-File Planning: Use Plan Mode to analyze a refactor or complex fix, review the proposed steps, and approve changes before execution.
Personal Assistant: Set up a local assistant that remembers project context and user preferences across tasks and resumed sessions.
Domain Agent Building: Develop a specialized agent by adding reusable Skills for recurring work and connecting external MCP tools when needed.
Agent Runtime Study: Explore a working Python implementation of model calls, tool execution, permission checks, context handling, and session recovery.

