{"id":47620424,"url":"https://github.com/sukethrp/agentos","last_synced_at":"2026-04-01T22:01:49.953Z","repository":{"id":336938153,"uuid":"1151135966","full_name":"sukethrp/agentos","owner":"sukethrp","description":"\"The Operating System for AI Agents. Build, Test, Deploy, Monitor, Govern.\"","archived":false,"fork":false,"pushed_at":"2026-03-21T21:30:36.000Z","size":845,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-03-22T07:42:41.590Z","etag":null,"topics":["agent-framework","agent-monitoring","agent-testing","agentos","ai-agent-platform","ai-agents","anthropic","fastapi","governance","llm","ollama","openai","python","rag","safety"],"latest_commit_sha":null,"homepage":"https://agentos-mocha.vercel.app","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/sukethrp.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":null,"code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-02-06T05:05:50.000Z","updated_at":"2026-03-21T21:30:39.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/sukethrp/agentos","commit_stats":null,"previous_names":["sukethrp/agentos"],"tags_count":3,"template":false,"template_full_name":null,"purl":"pkg:github/sukethrp/agentos","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sukethrp%2Fagentos","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sukethrp%2Fagentos/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sukethrp%2Fagentos/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sukethrp%2Fagentos/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sukethrp","download_url":"https://codeload.github.com/sukethrp/agentos/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sukethrp%2Fagentos/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31292631,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-01T21:15:39.731Z","status":"ssl_error","status_checked_at":"2026-04-01T21:15:34.046Z","response_time":53,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-framework","agent-monitoring","agent-testing","agentos","ai-agent-platform","ai-agents","anthropic","fastapi","governance","llm","ollama","openai","python","rag","safety"],"created_at":"2026-04-01T22:00:40.276Z","updated_at":"2026-04-01T22:01:49.930Z","avatar_url":"https://github.com/sukethrp.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cp align=\"center\"\u003e\n  \u003ch1 align=\"center\"\u003e🤖 AgentOS\u003c/h1\u003e\n  \u003cp align=\"center\"\u003e\u003cstrong\u003eThe Operating System for AI Agents\u003c/strong\u003e\u003c/p\u003e\n  \u003cp align=\"center\"\u003eBuild, Test, Deploy, Monitor, and Govern AI agents — from prototype to production.\u003c/p\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://opensource.org/licenses/Apache-2.0\"\u003e\u003cimg src=\"https://img.shields.io/badge/License-Apache_2.0-blue.svg\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://www.python.org/downloads/\"\u003e\u003cimg src=\"https://img.shields.io/badge/python-3.11+-blue.svg\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/sukethrp/agentos/actions\"\u003e\u003cimg src=\"https://github.com/sukethrp/agentos/actions/workflows/test.yml/badge.svg\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/sukethrp/agentos/releases\"\u003e\u003cimg src=\"https://img.shields.io/github/v/release/sukethrp/agentos\"\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://agentos-mocha.vercel.app\"\u003e🌐 Live Demo\u003c/a\u003e ·\n  \u003ca href=\"#quick-start\"\u003e🚀 Quick Start\u003c/a\u003e ·\n  \u003ca href=\"https://github.com/sukethrp/agentos/issues\"\u003e📋 Issues\u003c/a\u003e\n\u003c/p\u003e\n\n\u003c!-- Architecture diagram --\u003e\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"docs/assets/architecture.png\" alt=\"AgentOS Architecture\" width=\"700\"\u003e\n\u003c/p\u003e\n\n\u003e **For teams who need to deploy AI agents with testing, governance, and monitoring built in — not bolted on.**\n\n## 3 Differentiators\n\n- 🧪 **Test**: Run scenario-based simulation before deploy, with quality and cost scoring.\n- 🛡️ **Govern**: Enforce budgets, permissions, and kill-switch policies with auditability.\n- 📊 **Monitor**: Observe live agent runs, tool usage, latency, and spend in one dashboard.\n\n## Quick Start\n\n```bash\npip install agentos-platform\n```\n\n10-line example:\n\n```python\nfrom agentos.governed_agent import GovernedAgent\nfrom agentos.core.tool import tool\n\n@tool(description=\"Add two numbers\")\ndef add(a: float, b: float) -\u003e float:\n    return a + b\n\nagent = GovernedAgent(name=\"demo\", model=\"gpt-4o-mini\", tools=[add])\nprint(agent.run(\"What is 12.5 + 7.5?\"))\n```\n\nDemo mode:\n\n```bash\nAGENTOS_DEMO_MODE=true python examples/run_web_builder.py\n```\n\n## Features\n\n### MCP server with stdio/SSE transport (Claude Desktop + Cursor)\n\nInstall the MCP extra:\n\n```bash\npip install 'agentos-platform[mcp]'\n```\n\n### 1) Start the MCP server\n\nExpose built-in AgentOS tools (stdio transport is the safest choice for MCP clients like Claude Desktop and Cursor):\n\n```bash\nagentos mcp serve --transport stdio\n```\n\nExpose tools from a specific agent module (example `./my_agent/agent.py`):\n\n```bash\nagentos mcp serve --transport stdio --agent ./my_agent\n```\n\nOptional: run the HTTP SSE transport for clients that support it:\n\n```bash\nagentos mcp serve --transport sse --host 127.0.0.1 --port 8080\n```\n\n### 2) Configure Claude Desktop\n\nAdd the following snippet to your `claude_desktop_config.json` (restart Claude Desktop after editing):\n\n```json\n{\n  \"mcpServers\": {\n    \"agentos\": {\n      \"command\": \"agentos\",\n      \"args\": [\"mcp\", \"serve\", \"--transport\", \"stdio\"]\n    }\n  }\n}\n```\n\nIf you want a specific agent module:\n\n```json\n{\n  \"mcpServers\": {\n    \"agentos\": {\n      \"command\": \"agentos\",\n      \"args\": [\"mcp\", \"serve\", \"--transport\", \"stdio\", \"--agent\", \"/absolute/path/to/agent.py\"]\n    }\n  }\n}\n```\n\n### 3) Configure Cursor\n\nAdd to Cursor `.cursor/mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"agentos\": {\n      \"command\": \"agentos\",\n      \"args\": [\"mcp\", \"serve\", \"--transport\", \"stdio\"]\n    }\n  }\n}\n```\n\n### Agent delegation (delegate tool + SharedContext + chaining)\n\nAgentOS includes a structured delegation system that lets a “parent” agent offload subtasks to “child” agents while propagating rich context through a shared, in-memory key/value store.\n\nKey pieces:\n\n- `delegate_subtask` tool: LLM-facing tool that accepts structured fields like `task`, `context_json`, `constraints_json`, `expected_output_schema_json`, and `timeout`.\n- `SharedContext`: a key/value store child agents can read/write during the delegation chain (avoids lossy prompt compression).\n- Delegation chaining: if a child agent delegates again, the same shared context key is reused automatically.\n\nMinimal wiring example:\n\n```python\nfrom agentos.core.agent import Agent\nfrom agentos.core.delegation import DelegationManager\n\n# Define your child agents however you like.\nchild_agent_a = Agent(name=\"child-a\", model=\"gpt-4o-mini\", tools=[])\nchild_agent_b = Agent(name=\"child-b\", model=\"gpt-4o-mini\", tools=[])\n\nmanager = DelegationManager()\nmanager.register_agent(\"child-a\", child_agent_a)\nmanager.register_agent(\"child-b\", child_agent_b)\n\n# Create your parent agent and attach the delegate tool.\nparent = Agent(name=\"parent\", model=\"gpt-4o-mini\", tools=[])\nmanager.attach_delegate_tool(parent)  # adds `delegate_subtask` to the toolset\n\n# Now the parent agent can call `delegate_subtask`.\nparent.run(\"Delegate a subtask and use shared context for details.\")\n```\n\nSharedContext tools available to delegated agents:\n\n- `shared_context_key()`\n- `shared_context_get(key)`\n- `shared_context_set(key, value_json)`\n- `shared_context_dump()`\n\n## Core Modules\n\n| Module | What it does |\n|--------|---------------|\n| Agent SDK | Define agents and tools with provider-agnostic model routing |\n| Simulation Sandbox | Test scenarios with LLM-as-judge quality and pass/fail scoring |\n| Governance Engine | Budget controls, permissions, kill switch, and audit logging |\n| Live Dashboard | Real-time traces for prompts, tool calls, latency, and spend |\n| RAG Pipeline | Ingest, chunk, embed, and retrieve knowledge sources |\n| Workflow Engine | Compose repeatable multi-step agent workflows |\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cstrong\u003e📋 Full 15-module list (click to expand)\u003c/strong\u003e\u003c/summary\u003e\n\n| Module | Description |\n|--------|-------------|\n| Agent SDK | Core governed agent runtime and tool-calling loop |\n| WebSocket Streaming | Token streaming and low-latency interactive sessions |\n| RAG Pipeline | Ingestion, chunking, embeddings, retrieval, and reranking |\n| Simulation Sandbox | Scenario simulation, scoring, and comparison reports |\n| Live Dashboard | Event stream, usage analytics, and operational visibility |\n| Governance Engine | Guardrails, budget caps, permission checks, and audits |\n| Agent Scheduler | Interval and cron scheduling with execution history |\n| Event Bus | Trigger-driven orchestration via internal and external events |\n| Plugin System | Runtime-extensible tools, providers, and adapters |\n| Authentication | API key auth, org and user usage tracking, and middleware |\n| A/B Testing | Side-by-side evaluation for variants and prompt changes |\n| Workflow Engine | DAG-based execution with retries and branching |\n| Multimodal | Vision and document flows for image and file-aware agents |\n| Marketplace | Template registry for reusable agents and workflows |\n| Embed SDK | Embeddable widget and integration surface for web apps |\n\n\u003c/details\u003e\n\n## Honest Comparison\n\n| Capability | AgentOS | LangChain | CrewAI | AutoGen |\n|------------|---------|-----------|--------|---------|\n| Built-in testing sandbox | ✅ Native | ❌ External setup | ❌ External setup | ❌ External setup |\n| Governance (budget/kill switch) | ✅ Native | ⚠️ Custom code | ⚠️ Custom code | ⚠️ Custom code |\n| Real-time ops dashboard | ✅ Native | ⚠️ LangSmith add-on | ❌ | ❌ |\n| Batteries-included platform | ✅ Yes | ⚠️ Framework-first | ⚠️ Orchestration-first | ⚠️ Research-first |\n| Ecosystem maturity | 🌱 Growing | ✅ Very mature | ✅ Mature | ✅ Mature |\n\n## Benchmarks\nSee [full benchmark results](docs/benchmarks.md). Key findings:\n- Our weighted evaluation ensemble correlates 0.91 with human judgment\n- Local embeddings achieve 95% of OpenAI quality at zero cost\n- Governance adds \u003c5ms overhead to any query\n\n## Architecture\n\nSee the architecture diagram above and `docs/` for component-level details and ADRs.\n\n## Project Structure\n\n```text\nagentos/\n├── src/agentos/      # Core platform modules\n├── frontend/         # React frontend\n├── dashboard/        # Web dashboard UI\n├── deploy/helm/      # Helm charts\n├── examples/         # Runnable examples\n├── tests/            # Unit and integration tests\n└── docs/             # Docs and ADRs\n```\n\n## Contributing\n\nContributions are welcome: [CONTRIBUTING.md](CONTRIBUTING.md)\n\n## Roadmap\n\nRoadmap and upcoming work are tracked in [GitHub Issues](https://github.com/sukethrp/agentos/issues).\n\n- [ ] Agent-to-Agent mesh protocol\n- [x] MCP server with stdio/SSE transport\n- [x] Agent-to-agent delegation with shared context\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsukethrp%2Fagentos","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsukethrp%2Fagentos","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsukethrp%2Fagentos/lists"}