{"id":40482117,"url":"https://github.com/framersai/agentos","last_synced_at":"2026-05-10T08:14:42.238Z","repository":{"id":322156332,"uuid":"1084726333","full_name":"framersai/agentos","owner":"framersai","description":"Build autonomous AI agents with adaptive intelligence and emergent behaviors to simulate any scenario. Multimodal RAG included.","archived":false,"fork":false,"pushed_at":"2026-04-25T00:58:27.000Z","size":15351,"stargazers_count":77,"open_issues_count":0,"forks_count":3,"subscribers_count":1,"default_branch":"master","last_synced_at":"2026-04-25T02:41:06.311Z","etag":null,"topics":["agent-framework","agentic-ai","agentos","agents","ai-agents","anthropic","autonomous-agents","emergent-behavior","guardrails","hexaco","llm","multi-agent","openai","orchestration","personality","rag","tool-forging","typescript"],"latest_commit_sha":null,"homepage":"https://docs.agentos.sh","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/framersai.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":".github/CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":".github/CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":".github/SECURITY.md","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":"2025-10-28T04:35:14.000Z","updated_at":"2026-04-25T00:57:30.000Z","dependencies_parsed_at":"2026-02-19T02:01:40.284Z","dependency_job_id":"68a14ea1-6631-41e2-9737-36fd9fb948b3","html_url":"https://github.com/framersai/agentos","commit_stats":null,"previous_names":["wearetheframers/agentos","framersai/agentos"],"tags_count":283,"template":false,"template_full_name":null,"purl":"pkg:github/framersai/agentos","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/framersai%2Fagentos","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/framersai%2Fagentos/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/framersai%2Fagentos/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/framersai%2Fagentos/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/framersai","download_url":"https://codeload.github.com/framersai/agentos/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/framersai%2Fagentos/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32323215,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-26T23:26:28.701Z","status":"online","status_checked_at":"2026-04-27T02:00:06.769Z","response_time":128,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"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","agentic-ai","agentos","agents","ai-agents","anthropic","autonomous-agents","emergent-behavior","guardrails","hexaco","llm","multi-agent","openai","orchestration","personality","rag","tool-forging","typescript"],"created_at":"2026-01-20T18:33:39.584Z","updated_at":"2026-05-10T08:14:42.224Z","avatar_url":"https://github.com/framersai.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\n\u003ca href=\"https://agentos.sh\"\u003e\n  \u003cimg src=\"https://raw.githubusercontent.com/framersai/agentos/master/assets/agentos-primary-no-tagline-transparent-2x.png\" alt=\"AgentOS — TypeScript AI Agent Framework with Cognitive Memory\" height=\"100\" /\u003e\n\u003c/a\u003e\n\n\u003cbr /\u003e\n\n# **AgentOS** — Open-Source TypeScript AI Agent Runtime with Cognitive Memory, HEXACO Personality, and Runtime Tool Forging\n\n**85.6% on LongMemEval-S** at $0.0090/correct, +1.4 above Mastra OM gpt-4o (84.23%) · **70.2% on LongMemEval-M** (1.5M-token variant), the only open-source library on the public record above 65% on M with publicly reproducible methodology · 16 LLM providers · 8 neuroscience-backed memory mechanisms · Apache-2.0\n\n[![npm](https://img.shields.io/npm/v/@framers/agentos?style=flat-square\u0026logo=npm\u0026color=cb3837)](https://www.npmjs.com/package/@framers/agentos)\n[![CI](https://img.shields.io/github/actions/workflow/status/framersai/agentos/ci.yml?branch=master\u0026style=flat-square\u0026logo=github\u0026label=CI)](https://github.com/framersai/agentos/actions/workflows/ci.yml)\n[![tests](https://img.shields.io/badge/tests-3%2C866%2B_passed-2ea043?style=flat-square\u0026logo=vitest\u0026logoColor=white)](https://github.com/framersai/agentos/actions/workflows/ci.yml)\n[![codecov](https://codecov.io/gh/framersai/agentos/graph/badge.svg)](https://codecov.io/gh/framersai/agentos)\n[![TypeScript](https://img.shields.io/badge/TypeScript-5.4+-3178c6?style=flat-square\u0026logo=typescript\u0026logoColor=white)](https://www.typescriptlang.org/)\n[![License](https://img.shields.io/badge/License-Apache_2.0-blue?style=flat-square)](https://opensource.org/licenses/Apache-2.0)\n[![LongMemEval-S](https://img.shields.io/badge/LongMemEval--S-85.6%25-2ea043?style=flat-square)](https://docs.agentos.sh/blog/2026/04/27/longmemeval-s-83-with-semantic-embedder)\n[![LongMemEval-M](https://img.shields.io/badge/LongMemEval--M-70.2%25-2ea043?style=flat-square)](https://docs.agentos.sh/blog/2026/04/29/longmemeval-m-70-with-topk5)\n[![agentos-bench](https://img.shields.io/badge/bench-public-blue?style=flat-square)](https://github.com/framersai/agentos-bench)\n[![Discord](https://img.shields.io/badge/Discord-Join%20Us-5865F2?style=flat-square\u0026logo=discord)](https://wilds.ai/discord)\n\n[**Benchmarks**](https://github.com/framersai/agentos-bench/blob/master/results/LEADERBOARD.md) · [Website](https://agentos.sh) · [Docs](https://docs.agentos.sh) · [npm](https://www.npmjs.com/package/@framers/agentos) · [Discord](https://wilds.ai/discord) · [Blog](https://docs.agentos.sh/blog)\n\n\u003c/div\u003e\n\n---\n\nAgentOS is an open-source TypeScript runtime for AI agents that **remember, adapt, and write their own tools**.\n\nWhen an agent encounters a sub-task no existing tool covers, it generates a TypeScript function with a Zod-described schema, sends it through an LLM judge, and on approval runs it in a hardened `node:vm` sandbox. The new tool joins the catalog for the rest of the session. When a multi-agent team hits a capability gap, the manager calls `spawn_specialist` and the LLM judge reviews the synthesized agent spec before it joins the live roster.\n\nThe runtime carries the parts of an agent that should outlive a single chat completion: persistent [cognitive memory](https://docs.agentos.sh/features/cognitive-memory) (Ebbinghaus decay, retrieval-induced forgetting, reconsolidation, source-confidence decay) grounded in published cognitive-science literature, optional [HEXACO personality](https://docs.agentos.sh/features/cognitive-memory-guide) vectors that bias retrieval and routing, [six multi-agent orchestration strategies](https://docs.agentos.sh/features/multi-agent-collaboration), [streaming guardrails](https://docs.agentos.sh/features/guardrails-architecture), a [voice pipeline](https://docs.agentos.sh/features/voice-pipeline), and one dispatch interface across 21 LLM providers. Apache-2.0.\n\n[100+ first-party extensions](https://www.npmjs.com/package/@framers/agentos-extensions) (channel adapters, tool packs, guardrail packs) and [88 curated `SKILL.md` skills](https://www.npmjs.com/package/@framers/agentos-skills) auto-discover at startup through their respective registries: a host pulls a curated index and the runtime wires every tool, guardrail, channel, and skill without manual registration. The auto-loader is the same surface that runtime-forged tools join: an agent that invents a function in session N can promote it (with judge approval and `SkillExporter`) into a `SKILL.md` that the registry picks up on the next process start. Forging is how the surface grows mid-run; auto-discovery is how it ships as a first-class capability afterward.\n\nOn benchmarks: **85.6% on LongMemEval-S** at $0.0090 per correct answer (gpt-4o reader, +1.4 points above Mastra's published 84.23%, 0.4 points behind Emergence.ai's 86% closed-source SaaS SOTA); **70.2% on LongMemEval-M** (1.5M-token haystacks, 500 sessions per question), the only open-source library on the public record above 65% on M with publicly reproducible methodology. Per-case run JSONs and single-CLI reproduction ship in [agentos-bench](https://github.com/framersai/agentos-bench).\n\n---\n\n## Install\n\n```bash\nnpm install @framers/agentos\n```\n\n```typescript\nimport { agent } from '@framers/agentos';\n\nconst tutor = agent({\n  provider: 'anthropic',\n  instructions: 'You are a patient CS tutor.',\n  personality: { openness: 0.9, conscientiousness: 0.95 },\n  memory: { types: ['episodic', 'semantic'], working: { enabled: true } },\n});\n\nconst session = tutor.session('student-1');\nawait session.send('Explain recursion with an analogy.');\nawait session.send('Can you expand on that?'); // remembers context\n```\n\n[Full quickstart](https://docs.agentos.sh/getting-started) · [Examples cookbook](https://docs.agentos.sh/getting-started/examples) · [API reference](https://docs.agentos.sh/api)\n\n---\n\n## Emergent Design\n\n\u003e \"So we and our elaborately evolving computers may meet each other halfway.\"\n\u003e\n\u003e — Philip K. Dick, *The Android and the Human*, 1972\n\nThree things accumulate across an AgentOS session and compose into behavior:\n\n1. **Memory.** What was said, what was decided, what was retrieved.\n2. **Tool surface.** Starts at whatever was registered. Can grow when an agent forges a new function mid-decision and the judge approves it.\n3. **Personality** (optional). A HEXACO trait vector that biases retrieval, specialist routing, and decision-making.\n\nEach is configurable and observable; none crosses into \"emergent agent\" on its own. The composition is the interesting part.\n\n### Runtime Tool Forging\n\nWhen an agent encounters a sub-task that no available tool covers, it generates a TypeScript function with a Zod-described input and output schema. A separate LLM call evaluates the forged function against the agent's stated intent and either approves or rejects it. Approved functions execute in a hardened `node:vm` sandbox with strict defaults (5-second wall clock, 128 MB heap-delta budget, `eval` / `require` / `process` banned, `fetch` / `fs` / `crypto` allowlist-empty by default). Approved tools join a discoverable index keyed by name and signature; subsequent turns invoke them via `call_forged_tool(name, args)`. First forge costs full LLM tokens; reuse costs tens of tokens. Sandbox internals, isolation tradeoffs (`node:vm` vs queued `isolated-vm` for the hosted multi-tenant tier), and the full safety policy are in the [emergent capabilities docs](https://docs.agentos.sh/features/emergent-capabilities).\n\nThe pattern the runtime supports: an agent forges a tool mid-decision, the judge approves it, that turn invokes it, and a few turns later a different specialist agent in the same session invokes the same tool because the index made it findable. Promoted tools can be exported as `SKILL.md` skills via `SkillExporter` and join the auto-discovery surface on the next process start.\n\n### HEXACO Personality (optional)\n\nPersonality is opt-in. The runtime behaves identically with or without a trait vector, and most production deployments do not pass one.\n\n```ts\n// Personality-neutral (most production agents)\nconst support = agent({\n  provider: 'openai',\n  instructions: 'Resolve customer tickets.',\n  memory: { types: ['episodic', 'semantic'] },\n});\n\n// Opt-in HEXACO (when persona consistency across sessions matters)\nconst coach = agent({\n  provider: 'openai',\n  instructions: \"Long-running career coach. Hold the user accountable to their stated goals across weekly check-ins; flag drift, push back on excuses, escalate when goals shift.\",\n  personality: {\n    conscientiousness: 0.9,    // won't let goals drift between sessions\n    honestyHumility: 0.85,     // won't tell the user what they want to hear\n    emotionality: 0.3,         // stays steady when the user is reactive\n  },\n  memory: { types: ['episodic', 'semantic'] },\n});\n```\n\nWhen a vector is supplied, the kernel weights retrieval, specialist routing, and tool selection by the trait values. Same agent, same prompt, same tools: a high-Openness leader and a high-Conscientiousness leader produce measurably different decision sequences. Personality lives in the kernel, not in the prompt — prompt-only personality dissolves under context pressure while kernel-encoded bias persists. The vector remains editable, inspectable, and removable on consent.\n\n---\n\n## Memory Benchmarks\n\n`gpt-4o` reader, `gpt-4o-2024-08-06` judge, full N=500 across every row. Cross-provider numbers are excluded from the tables because their public methodology disclosures don't admit reproduction.\n\n### LongMemEval-S (115K tokens, 50 sessions)\n\n| System | Accuracy | $/correct | p50 latency |\n|---|---:|---:|---:|\n| EmergenceMem Internal | 86.0% | not published | 5,650 ms |\n| **AgentOS** (canonical-hybrid + reader-router) | **85.6%** | **$0.0090** | **3,558 ms** |\n| Mastra OM gpt-4o (gemini-flash observer) | 84.23% | not published | not published |\n| Supermemory gpt-4o | 81.6% | not published | not published |\n| EmergenceMem Simple Fast (rerun in agentos-bench) | 80.6% | $0.0586 | 3,703 ms |\n| Zep (self / independent reproduction) | 71.2% / 63.8% | not published | not published |\n\n+1.4 points above Mastra OM. EmergenceMem Internal posts 86.0% (0.4 above) but doesn't publish per-case results or a reproducible CLI; among open-source libraries with single-CLI reproduction at `gpt-4o`, 85.6% is the highest publicly reproducible number located. p50 latency 3,558 ms vs EmergenceMem's published median 5,650 ms.\n\nCross-provider numbers omitted from the table (different reader and/or undisclosed judge): Mastra OM 94.87% (gpt-5-mini + gemini-2.5-flash observer), agentmemory 96.2% (Claude Opus 4.6), MemMachine 93.0% (GPT-5-mini), Hindsight 91.4% (unspecified backbone).\n\n### LongMemEval-M (1.5M tokens, 500 sessions)\n\nM's haystacks exceed every production context window; most vendors only publish on S.\n\n| System | Accuracy | License |\n|---|---:|---|\n| LongMemEval paper, GPT-4o round Top-10 (paper's best) | 72.0% | open repo |\n| AgentBrain | 71.7% | closed-source SaaS |\n| LongMemEval paper, GPT-4o session Top-5 | 71.4% | open repo |\n| **AgentOS** (sem-embed + reader-router + Top-5) | **70.2%** | **Apache-2.0** |\n| LongMemEval paper, GPT-4o round Top-5 | 65.7% | open repo |\n| Mem0 v3, Mastra, Hindsight, Zep, EmergenceMem, Supermemory, Letta | not published | — |\n\nAt matched Top-5 retrieval, +4.5 above the round-level paper baseline (65.7%) and 1.2 below the session-level (71.4%); the paper's overall strongest GPT-4o result is 72.0% at Top-10. Of open-source libraries with publicly reproducible runs, AgentOS is the only one above 65% on M.\n\n\u003e **[Full leaderboard →](https://github.com/framersai/agentos-bench/blob/master/results/LEADERBOARD.md)** · **[Run JSONs →](https://github.com/framersai/agentos-bench/tree/master/results/runs)** · **[Transparency audit →](https://agentos.sh/en/blog/memory-benchmark-transparency-audit/)** · **[LongMemEval paper](https://arxiv.org/abs/2410.10813)** (Wu et al., ICLR 2025, Table 3)\n\nMethodology stack: bootstrap 95% CIs at 10k Mulberry32 resamples (seed 42), per-benchmark judge-FPR probes (S 1%, M 2%, LOCOMO 0%), per-case run JSONs, single-CLI reproduction. The [transparency audit](https://agentos.sh/en/blog/memory-benchmark-transparency-audit/) covers what the headline numbers don't: LOCOMO's ~6.4% answer-key error rate, the LongMemEval-S context-window confound, and the Mem0-vs-Zep comparison gaming case study, alongside which vendors disclose which methodology dimensions.\n\n---\n\n## Ecosystem\n\n| Package | Role |\n|---|---|\n| [`@framers/agentos`](https://www.npmjs.com/package/@framers/agentos) | Core runtime: GMI agents, cognitive memory, multi-agent orchestration, guardrails, voice, 21 LLM providers. Apache 2.0. |\n| [`@framers/agentos-extensions`](https://www.npmjs.com/package/@framers/agentos-extensions) | 100+ first-party extensions and templates: channel adapters, tool packs, integrations, guardrail packs. |\n| [`@framers/agentos-extensions-registry`](https://www.npmjs.com/package/@framers/agentos-extensions-registry) | Discovery + auto-loader layer for the extensions catalog. Hosts pull the index without pulling every implementation; the runtime resolves and registers packs at startup. |\n| [`@framers/agentos-skills`](https://www.npmjs.com/package/@framers/agentos-skills) | 88 curated `SKILL.md` skills covering common tasks. |\n| [`@framers/agentos-skills-registry`](https://www.npmjs.com/package/@framers/agentos-skills-registry) | Discovery + auto-loader layer for the skills catalog. Also the surface where promoted forged tools land after `SkillExporter`. |\n| [`@framers/agentos-bench`](https://github.com/framersai/agentos-bench) | Open benchmark harness. Bootstrap 95% CIs at 10k resamples, judge false-positive-rate probes, per-case run JSONs at fixed seed. MIT (the rest of AgentOS is Apache 2.0). |\n| [`@framers/sql-storage-adapter`](https://www.npmjs.com/package/@framers/sql-storage-adapter) | Cross-platform SQL persistence: SQLite, Postgres, IndexedDB, Capacitor SQLite. |\n| [`paracosm`](https://www.npmjs.com/package/paracosm) | AI agent swarm simulation engine that uses AgentOS as its substrate. |\n| [`wunderland`](https://www.npmjs.com/package/wunderland) | Sister project (preview) — batteries-included CLI plus daemon over the AgentOS extension and skill registries. 28-command CLI, 5-tier security, 8 agent presets, step-up HITL. [github.com/jddunn/wunderland](https://github.com/jddunn/wunderland). MIT. |\n\n**Extensions and skills auto-load at startup.** The runtime walks each registry plus any user-supplied paths, resolves each pack's `createExtensionPack(context)` factory or `SKILL.md` frontmatter, and registers tools, guardrails, channels, and skills without manual wiring. Capability gating and HITL approval gates apply to side-effecting installs. See [extensions architecture](https://docs.agentos.sh/architecture/extension-loading) for the full loading model.\n\n---\n\n## 📄 Technical Whitepaper · Coming Soon\n\nThe full architecture and benchmark methodology, written for engineers and researchers who want a citable PDF instead of scrolling docs. Cognitive memory pipeline, classifier-driven dispatch, HEXACO personality modulation, runtime tool forging, full LongMemEval-S/M and LOCOMO benchmark methodology with confidence interval math, judge-FPR probes, per-stage retention metrics, and reproducibility recipes.\n\n| Covers | What's inside |\n|---|---|\n| **Architecture** | Generalized Mind Instances, IngestRouter / MemoryRouter / ReadRouter, 8 cognitive mechanisms with primary-source citations |\n| **Benchmarks** | LongMemEval-S 85.6%, LongMemEval-M 70.2%, vendor landscape, confidence interval methodology, judge FPR probes, full transparency stack |\n| **Reproducibility** | Per-case run JSONs at `--seed 42`, single-CLI reproduction, Apache-2.0 bench at [github.com/framersai/agentos-bench](https://github.com/framersai/agentos-bench) |\n\n**[Join Discord for the announcement →](https://wilds.ai/discord)** · **[Read the benchmarks now →](https://github.com/framersai/agentos-bench/blob/master/results/LEADERBOARD.md)**\n\n---\n\n## Classifier-Driven Memory Pipeline\n\nMost memory libraries retrieve on every query. AgentOS gates memory through three LLM-as-judge classifiers in a single shared pass, so trivial queries skip retrieval entirely and the rest get the right architecture and reader per category.\n\n```\nUser query\n    │\n    ▼ Stage 1: QueryClassifier (gpt-5-mini, ~$0.0001/query)\n    │    T0=none ─────► answer from context, skip retrieval\n    │    T1+=needs memory\n    ▼ Stage 2: MemoryRouter      → canonical-hybrid · OM-v10 · OM-v11\n    ▼ Stage 3: ReaderRouter      → gpt-4o (TR/SSU) · gpt-5-mini (SSA/SSP/KU/MS)\n    ▼\nGrounded answer\n```\n\nStages 2 and 3 reuse the Stage 1 classification, so the full pipeline costs **one classifier call per query**, not three. **The T0 / no-memory gate is the novel piece**: removing retrieval entirely for greetings and small talk saves the embedding + rerank + reader cost on a substantial fraction of typical agent traffic.\n\n| Primitive | Source | Decision |\n|---|---|---|\n| `QueryClassifier` | [`@framers/agentos/query-router`](https://docs.agentos.sh/features/query-routing) | T0/none vs T1/simple vs T2/moderate vs T3/complex |\n| `MemoryRouter` | [`@framers/agentos/memory-router`](https://docs.agentos.sh/features/memory-router) | canonical-hybrid vs observational-memory-v10 vs v11 |\n| `ReaderRouter` | [`@framers/agentos/memory-router`](https://docs.agentos.sh/features/memory-router) | gpt-4o vs gpt-5-mini per category |\n\n[Cognitive Pipeline docs →](https://docs.agentos.sh/features/cognitive-pipeline) · [Architecture deep dive →](https://docs.agentos.sh/blog/2026/04/10/cognitive-memory-architecture-deep-dive) · [Beyond RAG →](https://docs.agentos.sh/blog/2026/03/31/cognitive-memory-beyond-rag)\n\n---\n\n## Why AgentOS\n\n| vs. | AgentOS differentiator |\n|---|---|\n| **LangChain / LangGraph** | Cognitive memory ([8 neuroscience-backed mechanisms](https://docs.agentos.sh/features/cognitive-memory)), HEXACO personality, runtime tool forging |\n| **Vercel AI SDK** | Multi-agent teams (6 strategies), 7 vector backends, [guardrails](https://docs.agentos.sh/features/guardrails-architecture), voice/telephony |\n| **CrewAI / Mastra** | Unified orchestration (DAGs + graphs + missions), personality-driven routing, **published reproducible numbers on LongMemEval-S (85.6%) and LongMemEval-M (70.2%) with full methodology disclosure** |\n\n[Full framework comparison →](https://docs.agentos.sh/blog/2026/02/20/agentos-vs-langgraph-vs-crewai)\n\n---\n\n## Key Features\n\n| Category | Highlights |\n|---|---|\n| **LLM Providers** | 16: OpenAI, Anthropic, Gemini, Groq, Ollama, OpenRouter, Together, Mistral, xAI, Claude/Gemini CLI, + 5 image/video |\n| **Cognitive Memory** | 8 mechanisms: reconsolidation, retrieval-induced forgetting, involuntary recall, FOK, gist extraction, schema encoding, source decay, emotion regulation |\n| **HEXACO Personality** | 6 traits modulate memory, retrieval bias, response style |\n| **RAG Pipeline** | 7 vector backends · 4 retrieval strategies · GraphRAG · HyDE · Cohere rerank-v3.5 |\n| **Multi-Agent Teams** | 6 coordination strategies · shared memory · inter-agent messaging · HITL gates |\n| **Orchestration** | `workflow()` DAGs · `AgentGraph` cycles · `mission()` goal-driven planning · checkpointing |\n| **Guardrails** | 5 security tiers · 6 packs (PII, ML classifiers, topicality, code safety, grounding, content policy) |\n| **Emergent Capabilities** | Runtime tool forging · 4 self-improvement tools · tiered promotion · skill export |\n| **Voice \u0026 Telephony** | ElevenLabs, Deepgram, Whisper · Twilio, Telnyx, Plivo |\n| **Channels** | 37 platform adapters (Telegram, Discord, Slack, WhatsApp, webchat, ...) |\n| **Observability** | OpenTelemetry · usage ledger · cost guard · circuit breaker |\n\n---\n\n## Multi-Agent in 6 Lines\n\n```typescript\nimport { agency } from '@framers/agentos';\n\nconst team = agency({\n  strategy: 'graph',\n  agents: {\n    researcher: { provider: 'anthropic', instructions: 'Find relevant facts.' },\n    writer:     { provider: 'openai',    instructions: 'Summarize clearly.',  dependsOn: ['researcher'] },\n    reviewer:   { provider: 'gemini',    instructions: 'Check accuracy.',     dependsOn: ['writer'] },\n  },\n});\n\nconst result = await team.generate('Compare TCP vs UDP for game networking.');\n```\n\nStrategies: `sequential` · `parallel` · `debate` · `review-loop` · `hierarchical` · `graph`. With `strategy: 'hierarchical'` + `emergent: { enabled: true }`, the manager LLM gets a `spawn_specialist` tool that mints new sub-agents at runtime when the static roster doesn't cover a sub-task. `agency()` is for single-request multi-agent coordination — for long-running world simulations or per-turn parallel agent loops, build your own orchestration with `agent()` + the lower-level primitives. [Multi-agent docs →](https://docs.agentos.sh/features/multi-agent) · [Hierarchical + emergent →](https://docs.agentos.sh/architecture/emergent-agency-system) · [Scope guide →](https://docs.agentos.sh/orchestration/agency-api#scope-when-to-reach-for-agency)\n\n---\n\n## See It In Action\n\n### 🌀 Paracosm — AI Agent Swarm Simulation\n\nDefine any scenario as JSON. Run it with AI commanders that have different HEXACO personalities. Same starting conditions, different decisions, divergent civilizations. Built on AgentOS.\n\n```bash\nnpm install paracosm\n```\n\n[Live Demo](https://paracosm.agentos.sh/sim) · [GitHub](https://github.com/framersai/paracosm) · [npm](https://www.npmjs.com/package/paracosm)\n\n---\n\n## Configure API Keys\n\nThree layers, highest priority first:\n\n```typescript\n// 1. Inline on the call (per-tenant, per-test, per-customer)\ngenerateText({ apiKey: 'sk-customer', prompt: '...' });\n\n// 2. Module-level default — set once at boot, no .env needed\nimport { setDefaultProvider } from '@framers/agentos';\nsetDefaultProvider({ provider: 'openai', apiKey: process.env.MY_OWN_KEY });\n\n// 2b. Reorder the env-var auto-detect chain instead (when you keep multiple keys)\nimport { setProviderPriority } from '@framers/agentos';\nsetProviderPriority(['anthropic', 'openai', 'ollama']);\n\n// 3. Environment variable auto-detect chain (default order)\n//    OpenRouter → OpenAI → Anthropic → Gemini → Groq → Together → Mistral\n//    → xAI → claude CLI → gemini CLI → Ollama → image providers\n```\n\n```bash\nexport OPENAI_API_KEY=sk-...\nexport ANTHROPIC_API_KEY=sk-ant-...\nexport GEMINI_API_KEY=AIza...\n\n# Comma-separated keys auto-rotate with quota detection\nexport OPENAI_API_KEY=sk-key1,sk-key2,sk-key3\n```\n\n[Full credential resolution + default models per provider →](https://docs.agentos.sh/architecture/llm-providers)\n\n---\n\n## API Surfaces\n\n- **`agent()`**: lightweight stateful agent. Prompts, sessions, personality, hooks, tools, memory.\n- **`agency()`**: multi-agent teams + full runtime. Emergent tooling, guardrails, RAG, voice, channels, HITL.\n- **`generateText()` / `streamText()` / `generateObject()` / `generateImage()` / `generateVideo()` / `generateMusic()` / `performOCR()` / `embedText()`**: low-level multi-modal helpers with native tool calling.\n- **`workflow()` / `AgentGraph` / `mission()`**: three orchestration authoring APIs over one graph runtime.\n\nProvider fallback is an explicit opt-in via `agent({ fallbackProviders: [...] })` (or `buildFallbackChain()` for programmatic chains). Defaults to off — the runtime never silently retries against a different provider unless you configured a chain.\n\n[Full API reference →](https://docs.agentos.sh/api) · [High-Level API guide →](https://docs.agentos.sh/getting-started/high-level-api)\n\n---\n\n## Documentation \u0026 Community\n\n- **[Benchmarks](https://github.com/framersai/agentos-bench/blob/master/results/LEADERBOARD.md)**: benchmark tables, 95% confidence intervals, methodology audit\n- **[Architecture](https://docs.agentos.sh/architecture/system-architecture)**: system design, layer breakdown\n- **[Cognitive Memory](https://docs.agentos.sh/features/cognitive-memory)**: 8 mechanisms with 30+ APA citations\n- **[RAG Configuration](https://docs.agentos.sh/features/rag-memory-configuration)**: vector stores, embeddings, sources\n- **[Guardrails](https://docs.agentos.sh/features/guardrails-architecture)**: 5 tiers, 6 packs\n- **[Voice Pipeline](https://docs.agentos.sh/features/voice-pipeline)**: TTS, STT, telephony\n- **[Blog](https://docs.agentos.sh/blog)**: engineering posts, benchmark publications, transparency audits\n- **[Discord](https://wilds.ai/discord)** · **[GitHub Issues](https://github.com/framersai/agentos/issues)** · **[Wilds.ai](https://wilds.ai)** (AI game worlds powered by AgentOS)\n\n---\n\n## Contributing\n\n```bash\ngit clone https://github.com/framersai/agentos.git \u0026\u0026 cd agentos\npnpm install \u0026\u0026 pnpm build \u0026\u0026 pnpm test\n```\n\n[Contributing Guide](https://github.com/framersai/agentos/blob/master/CONTRIBUTING.md) · We use [Conventional Commits](https://www.conventionalcommits.org/).\n\n---\n\n## License\n\n[Apache 2.0](./LICENSE)\n\n\u003cdiv align=\"center\"\u003e\n\n\u003ca href=\"https://agentos.sh\"\u003e\n  \u003cimg src=\"https://raw.githubusercontent.com/framersai/agentos/master/assets/agentos-primary-transparent-2x.png\" alt=\"AgentOS\" height=\"40\" /\u003e\n\u003c/a\u003e\n\u0026nbsp;\u0026nbsp;\u0026nbsp;\n\u003ca href=\"https://frame.dev\"\u003e\n  \u003cimg src=\"https://raw.githubusercontent.com/framersai/agentos/master/assets/frame-logo-green-no-tagline.svg\" alt=\"Frame.dev\" height=\"40\" /\u003e\n\u003c/a\u003e\n\n**Built by [Manic Agency LLC](https://manic.agency) · [Frame.dev](https://frame.dev) · [Wilds.ai](https://wilds.ai)**\n\n\u003c/div\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fframersai%2Fagentos","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fframersai%2Fagentos","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fframersai%2Fagentos/lists"}