Agent Framework Source Notes
Agent Framework Source Notes
This series is not a framework ranking. It reads frameworks by the runtime owners they introduce: model loop, tool runtime, state ledger, permission gate, workflow graph, service plane, and multi-agent coordination. The route starts with a shared map, then uses Pi to fix a minimal core, AgentScope to expand the runtime, a worksite chapter to separate Coding / General from Local / Cloud, and a protocol chapter to draw system boundaries before moving through ADK, Agno, AutoGen, CrewAI, and Eino, then closing with three tRPC-Agent-Go chapters on context, Memory, and Evolution.
Source claims come from public repository snapshots, while product positioning comes from project READMEs and official docs. Each later chapter follows one request or task in execution order.
Use the Claude Code and Codex runtime routes as a reference point, then compare general-purpose agent frameworks by the responsibilities they own.
Source Notes · Part 2 Why Pi Treats A Coding Agent As An Extensible Harness
Follow pi-ai, agent-core, and coding-agent to see how a minimal runtime owns the tool loop, session tree, and compaction while leaving plans, sub-agents, MCP, and permission policy to extensions.
Source Notes · Part 3 Following AgentScope reply_stream
Expand Pi's minimal baseline by following AgentScope through events, state, tool permissions, and the real adapter difference between OpenAI Responses and Chat Completions.
Source Notes · Part 4 Where Does the Agent Work?
Use one scripting task to separate Coding / General from Local / Cloud, then identify the owners of workspace, sandbox, session, artifact, and memory.
Source Notes · Part 5 Who Does An Agent Protocol Connect?
Use one contract-renewal task to separate the boundaries owned by MCP, A2A, AG-UI, Agent Client Protocol, and model Provider APIs.
Source Notes · Part 6 Why ADK Python Puts Agent And Workflow Together
Read how ADK 2.0 combines autonomous agents, deterministic workflow, Runner, session events, and Task API in one code-first runtime.
Source Notes · Part 7 Agno Turns Agents Into A Platform, Not Just Objects
Read how Agno uses AgentOS to connect agents, teams, workflows, APIs, storage, approvals, RBAC, scheduler, and interfaces into one control plane.
Source Notes · Part 8 From AutoGen To Microsoft Agent Framework
AutoGen is now in maintenance mode. Read from Core runtime and AgentChat teams to MAF Agent, Workflow, Orchestrations, and Hosting.
Source Notes · Part 9 CrewAI Models Agents As Team Workflow
Read Agent, Task, Crew, Process, and Flow source to see how CrewAI separates team autonomy from production control.
Source Notes · Part 10 Eino Turns Agent Apps Into Composable Go Execution Graphs
Read Component, Runnable, compose Graph, checkpointing, callbacks, ADK Runner, ChatModelAgent, and AgentTool as Eino's typed Go runtime.
Source Notes · Part 11 tRPC-Agent-Go Runner And Context Governance
Read the full Runner, Agent, Model, and service runtime, then follow Summary, Context Compaction, and Session Recall through compression, rebuild, and evidence recovery.
Source Notes · Part 12 How tRPC-Agent-Go Memory Turns Experience Into Searchable Knowledge
Follow one environment fact through delta, extraction, reconciliation, Fact/Episode storage, and hybrid retrieval, then test quality, cost, and refusal regressions on LoCoMo-10.
Source Notes · Part 13 How tRPC-Agent-Go Evolution Learns And Rejects Bad Skills
Read post-run review, revisions, and gates for online Evolution, then use SkillCraft and GEPA-inspired optimization to separate discovery, disconfirmation, submission, and release governance.