Agentic AI Ecosystem Map

A hand-curated taxonomy of the agentic AI ecosystem — categories, sub-areas, and representative projects as evidence that each functional slot exists. This is a classification map, not a project directory.

• Full taxonomy tree: SCHEMA.md (generated)
• Source of truth: graph/nodes.json + graph/edges.json
• Derived tree view: ecosystem.json
• Interactive 3D viz: danielrosehill.github.io/Agentic-AI-Ecosystem/viz/
• Sister project — runtime architecture model: Agentic-AI-Architecture-Visualisation

At a glance

11 top-level clusters · 435 sub-areas · 307 example projects.

Clusters

Each row is a top-level functional cluster. Sub-areas and examples are truncated — see SCHEMA.md for the complete tree.

Cluster	What it covers	Sub-areas	Example projects
Agents	Agent runtimes, frameworks, harnesses, and surfaces that define, host, or coordinate individual and multi-agent systems.	Agent Clusters · Agent Collaboration · Agent to Agent · Autonomous · Autonomous Agent Creation · Backend Platforms · +25	Xata Agent, IntentKit, AgentVerse, AgentChattr, +37
Builders & Platforms	Platforms and products for building, deploying, and operating agents — the IDE and PaaS layer of the ecosystem.	Automation · Deployment Platforms · Development Platforms · Engineering Platforms · On Device Agents · Optimisation Platform · +2	Activepieces, Flowise, Julep, Coze Studio, +8
Context, Memory & Data	Everything that feeds agents with context, persistent memory, and structured data.	Context · Data · Memory	PromptX, WeKnora, context-mode, Contexto, +40
Dev Tools & Skills	Tooling, reusable skills, and prompt assets that developers use to build and operate agents.	Dev Tools · Prompts · Skills	Aider, OpenDevin/OpenHands, SWE-agent, vLLora, +19
Ecosystem	Meta-layer covering non-tool entities in the agentic AI landscape: organisations, destinations, and uncategorised items.	Destinations · Organisations · Publications & Newsletters · Uncategorized	Agentic AI Industry Foundation (AAIF), Agency Agents
Frameworks & Runtimes	Top-level domain for the software substrate that executes agents: frameworks, runtimes, harnesses, loops, and orchestration.	Environments · Frameworks · Harnesses · Loops · Orchestration · Pipelines · +2	agent-os, Abacus Desktop, cua.ai, AgentForge, +48
Interfaces	User-facing surfaces through which humans interact with agents — frontends, workspaces, UX patterns, and annotation tools.	Annotation · Control Surfaces · Frontends · UX · Workspaces	Operit, Agent UI, Goose, Agent Deck, +5
Models & Inference	The model layer: base models themselves and the infrastructure that serves them to agents.	Inference · Models	Context-Gateway, LMDeploy, SGLang, Text Generation Inference, +27
Observability & Evals	Telemetry, evaluation, and operational monitoring for LLM and agent systems across their lifecycle.	Agent Analytics · Benchmarks · Evals and Experiment Tracking · Operations · Research & Experimentation	RagaAI Catalyst, AICGSecEval, OSWorld, AgentBench, +22
Safety & Governance	Controls, policies, and oversight mechanisms that keep agents safe, compliant, and accountable in production.	Agent in the Loop · Guardrails · Hallucination & Factuality · Human in the Loop · Identity and Auth · PII and Redaction · +3	Shellfirm, Guardrails AI, Invariant Labs, Llama Guard, +14
Tools & Integrations	Tools, actions, and integration layers that connect agents to the outside world — MCP, standards, and action surfaces.	Actions · MCP · Standards	BrowserOS, Lightpanda Browser, Steel Browser, Arcade.dev, +45

---

README generated from ecosystem.json by scripts/generate_readme.py. Full tree generated into SCHEMA.md by scripts/generate_schema.py.