The Agentic Map
Agents are the visible part of a much larger system. The Agentic Map organises that system into the layers a company must understand before it can deploy agents with confidence.
This is a living structure. Categories will be revised as evidence accumulates and the market resolves.
01
Models, inference providers and routing systems that supply the reasoning capacity behind agent work.
Includes Model access, compute markets, routing, failover, latency optimisation, cost management and workload-based model selection.
02
The environments in which agents plan, execute, maintain state and coordinate with other agents.
Includes Agent runtimes, memory, workflow orchestration, multi-agent coordination, deployment, sandboxes and version control.
03
The connection layer through which agents securely reach software, data and external systems.
Includes MCP infrastructure, connectors, context systems, secrets management, browser and computer use, data permissions and provenance.
04
The systems that establish what an agent is, who it represents and which credentials can be trusted.
Includes Agent identity, organisational binding, verifiable credentials, attestation, KYA, KYC and KYB connections.
05
The policy layer defining what an agent may see, decide, spend, change or commit to on behalf of its principal.
Includes Delegated authority, approval limits, consent, revocation, separation of duties, escalation rules and kill switches.
06
Protection against malicious instructions, compromised tools, data leakage and unsafe agent behaviour.
Includes Prompt-injection defence, tool-supply-chain security, behavioural monitoring, red-teaming, isolation, recovery and incident response.
07
The systems used to measure what agents do, how reliably they do it and when their behaviour changes.
Includes Testing, simulation, benchmarks, trace analysis, continuous evaluation, cost measurement, quality monitoring and forensic replay.
08
The mechanisms through which companies and agents find, assess, approve and purchase external agent capabilities.
Includes Registries, catalogues, capability search, marketplaces, vendor risk, agent procurement and task routing.
09
The commercial layer that defines the work, records acceptance and determines whether the promised outcome occurred.
Includes Machine-readable scopes, service levels, mandates, contracting, acceptance criteria, outcome verification and dispute evidence.
10
The systems enabling bounded spending, settlement, billing, credit and financial reconciliation for agent activity.
Includes Payment mandates, wallets, spend controls, metering, outcome billing, stablecoin and card abstraction, credit, tax and accounting.
11
The controls and records that make agent deployment intelligible to management, boards, auditors and regulators.
Includes Agent inventories, ownership, policy management, audit trails, decision provenance, compliance evidence, rights and recourse.
12
The operating systems through which people approve, supervise, correct and collaborate with fleets of agents.
Includes Exception queues, approval interfaces, escalation, role design, agent workforce management, capacity planning and performance management.
Companies and initiatives
Classification uses public primary sources. Inclusion is not testing, endorsement or verification, and vendor descriptions are labelled as claims.
Models and APIs used as intelligence components inside agent systems.
Claude models and developer platform used for tool-using and agentic applications.
Models and APIs available through Google AI and Vertex AI for agent development.
Managed access to foundation models and services for building generative-AI applications and agents.
Microsoft’s platform for building, evaluating and managing AI applications and agents.
Model provider offering commercial and open-weight models through APIs and deployment options.
Enterprise model platform focused on language, retrieval and tool-using applications.
Inference, fine-tuning and model infrastructure for open and custom models.
Inference and model-serving platform for generative-AI workloads.
Inference infrastructure built around Groq’s language-processing hardware and cloud service.
Graph-based orchestration for long-running, stateful agent workflows within the LangChain ecosystem.
Framework and platform for building coordinated groups of specialised agents and workflows.
Framework and platform for agents operating over enterprise data and tools.
Open-source framework for developing multi-agent applications.
Google’s framework for developing and orchestrating agents.
Developer toolkit for tool use, hand-offs, guardrails, sessions and tracing in agent applications.
Open-source SDK for model-driven agent applications in the AWS ecosystem.
Python agent framework built around typed application development.
TypeScript framework for agents, workflows, retrieval, memory and evaluation.
Microsoft’s open-source SDK for integrating models, plugins and agent orchestration.
Open protocol for connecting AI applications with external tools and data sources.
Tool-integration and authentication platform for AI agents.
Automation platform exposing application actions and agent-oriented products across a large integration ecosystem.
Developer integration platform for connecting APIs, data sources and workflows, including agent tools.
Enterprise automation and integration platform adding agent-oriented orchestration and governance.
Unified APIs for connecting products with multiple business-software categories.
Browser infrastructure for agents and automated web workflows.
Web-data extraction and crawling tools designed for AI applications.
Tool-calling and authorisation infrastructure for agents.
Developer and security infrastructure including browser rendering, AI Gateway and remote MCP-related services.
Enterprise identity and access platform increasingly relevant to non-human identities and agent access.
Identity platform covering human and workload identities within Microsoft environments.
Customer and application identity platform used to authenticate users and machine-to-machine applications.
Enterprise authentication, directory and identity infrastructure for software products.
Identity and authentication infrastructure for users, organisations and machine-to-machine access.
Identity platform for authentication, authorisation and application access.
Workload identity and access management for non-human access between applications and services.
Continuous and dynamic access-management infrastructure.
Identity and knowledge-graph infrastructure for digital identities and relationships.
Standardised self-description of an agent’s capabilities and interaction requirements; not, on its own, verification of trustworthiness.
Fine-grained authorisation platform for application and service permissions.
Authorisation-as-a-service and policy infrastructure for application permissions.
Enterprise authorisation platform for centrally managed access policy.
Open-source and managed authorisation layer for application permissions.
General-purpose open-source policy engine used to decouple policy decisions from application logic.
Permissions infrastructure based on relationship-based access control.
Authorisation infrastructure and policy tooling for applications.
Access infrastructure for people and services interacting with systems and resources.
Identity-security platform covering privileged access and machine identities.
Secrets and identity-based access infrastructure frequently used for application and workload credentials.
Positions an AI security platform covering discovery, posture, governance and runtime protection for models and agents.
AI security portfolio including runtime protection and red-team capabilities built in part on Lakera.
Security controls for enterprise use of generative AI applications and agents.
AI security platform focused on discovery, testing and protection of AI applications and agents.
Security platform for machine-learning and AI systems.
AI/ML security platform and open-source security ecosystem.
AI security testing and red-team platform.
Security platform for generative-AI applications, data and agent interactions.
Enterprise AI security, testing and runtime control platform.
Enterprise AI security capabilities covering development and use of AI applications.
Platform for tracing, evaluating, testing and monitoring LLM and agent applications.
Evaluation and observability platform for AI products.
AI engineering and observability platform for evaluating and improving production systems and agents.
Open-source LLM engineering platform covering traces, evaluation, prompts and datasets.
Toolkit and platform for tracking, evaluating and improving generative-AI applications.
Observability, testing and monitoring focused on agent development and production operation.
Evaluation and testing platform for AI systems.
Testing and evaluation tools for AI models and applications.
Open-source evaluation framework and commercial platform for LLM systems.
Open-source platform including tracing and evaluation capabilities for generative-AI applications and agents.
Agent Cards describe capabilities; discovery methods vary by environment and trust requirement.
Registry infrastructure for discovering publicly available MCP servers.
Public catalogue for discovering payment-gated services indexed through the x402 ecosystem.
Cloud marketplace surface for discovering partner-built AI agents.
Marketplace category for third-party agent products and tools deployable through AWS.
Marketplace and partner ecosystem for Agentforce-related actions, topics and agent components.
Marketplace distribution for software and agent-related solutions in the Microsoft ecosystem.
Open ecosystem for models, datasets and applications with growing relevance to agent discovery and reusable components.
Code, package and project discovery surface increasingly important to MCP servers, agent frameworks and open tools.
Tool and integration catalogue oriented towards agent access across software applications.
Agreement-management and identity products relevant to human- and software-assisted contracting.
Digital contracting platform with workflow, repository and AI capabilities.
Enterprise contract-intelligence and lifecycle-management platform.
Contract creation, approval, signing and repository platform with AI features.
Contract lifecycle-management platform.
Contract-management platform focused on obligations and performance.
Legal and contract-management platform with AI capabilities.
Positions payment infrastructure for AI agents, including metering and monetisation of agent services.
Positions identity and payment infrastructure for agent transactions.
Open protocol constructs intended to carry evidence of user intent and transaction authorisation.
Merchant infrastructure for product discovery, checkout and payments through agents across supported protocols.
Open protocol for secure, accountable agent payments using mandates and existing payment methods.
Open HTTP-native protocol for request-level payment used by people, software and agents.
Payment-network initiative covering trusted agent credentials, permissioning and transaction flows.
Visa initiative for enabling AI agents in commerce through tokenisation, identity and payment-network controls.
Payments platform building merchant and agent-commerce capabilities across major commerce ecosystems.
Agent identity and payment platform for enabling agents to pay and receive payment.
Payment, metering and monetisation infrastructure for AI agents and services.
Positions an API through which agents can make payments to people and businesses with controls.
Stablecoin and wallet infrastructure relevant to programmable and agent-initiated settlement.
Centralised management and governance for enterprise AI agents, models and workflows in the ServiceNow ecosystem.
Governance, risk and compliance tooling for AI models and applications.
Data governance, compliance and risk capabilities extending across AI use in Microsoft environments.
AI governance, risk and compliance platform.
AI governance, risk management and compliance platform and services.
Enterprise AI governance platform for inventories, controls and reporting.
AI governance and compliance platform.
Governance capabilities linking AI use with data and policy management.
AI governance and risk-management capabilities within OneTrust.
Governance and oversight for analytics and AI projects within the Dataiku platform.
Workday initiative for managing AI agents alongside workforce roles and business processes.
Administrative, security and management controls for Copilot and agents in Microsoft environments.
Platform for building and operating agents inside customer workflows, including human hand-off.
Orchestration and supervision of teams of agents within enterprise workflows.
Enterprise platform for building, managing and governing agent-based work.
Enterprise agents operating across connected workplace knowledge and systems.
AI copilot and agent capabilities embedded across SAP business applications and processes.
Automation platform combining robots, agents and human-in-the-loop workflow.
Enterprise automation platform adding agentic-process and human-supervision capabilities.
Platform for building and operating teams of specialist agents with human oversight.
Help shape the map
We are compiling company profiles across every category. Submissions are researched independently; inclusion is never sold and never guaranteed.