Published records
39 tools
One record is counted for each published tool in the live directory index.
Research snapshot · Data current to 6 September 2026 UTC
State of Agent-First Infrastructure
A live count of the directory's published records, with an explanation of what those records can show and what they cannot show. This is a taxonomy snapshot, not a ranking, adoption survey, or estimate of the whole market.
Active taxonomy
14 categories
Only indexable categories containing at least one published, indexable listing are counted.
Required classification
3 classes
Agent-native, agent-enabling, and agent internet protocol describe why a listing qualifies.
Classification distribution
How published listings qualify
Agent-native22 listings
Agent-enabling15 listings
Agent internet protocol2 listings
Pricing labels
How listings describe access
Free2 listings
Freemium13 listings
Open Source11 listings
Paid8 listings
Pricing not documented2 listings
Source Available3 listings
These are broad editorial labels, not a price comparison. Plans, limits, licences, and usage charges can change; confirm current terms with the provider.
Category coverage
What each category means
5 published tools
Identity & Comms
Identity and communication tools belong here when they are designed for agents to be addressed, authenticated, contacted, or represented in an ongoing workflow.
Include when: An agent can provision or operate its own inbox, number, account, or communication identity.
Exclude when: It is a conventional chat, CRM, or telecom product with only a generic API.
2 published tools
Compute & Sandboxes
Compute tools belong here when they materially improve an agent’s ability to execute, inspect, resume, or contain code, command-line, desktop, or remote-machine work.
Include when: Agents can provision, control, or resume an isolated execution environment programmatically.
Exclude when: It is ordinary cloud hosting or container infrastructure with no substantive agent workflow.
4 published tools
Browser Automation
Browser automation tools belong here when their primary value is operating a browser or website interface for an agent, from low-level browser control to higher-level task execution.
Include when: The agent can navigate, click, type, inspect, or extract through a real or emulated browser.
Exclude when: Its main purpose is crawling or structured extraction without browser interaction.
3 published tools
UI & Frontends
Frontend systems belong here when they are built around agent-generated interfaces, agent-user interaction, or shared human-agent work rather than conventional application UI alone.
Include when: Agents can create or materially edit the interface or its content.
Exclude when: It is a general component library or frontend framework with no agent-specific capability.
2 published tools
Agent Testing & QA
Testing and quality-assurance systems belong here when agents actively author, operate, evaluate, or debug tests.
Include when: An agent can author, execute, adapt, or diagnose tests through a documented interface.
Exclude when: It is a conventional test runner that an agent can merely invoke.
2 published tools
Storage & Media
Storage and media infrastructure belongs here when it has a documented agent-first workflow, such as direct tool access, durable artifact URLs, transformations, or agent-oriented discovery.
Include when: Agents can store, retrieve, transform, or share artifacts through a documented machine interface.
Exclude when: It is conventional file hosting or object storage with no substantive agent workflow.
3 published tools
Memory & State
Memory and state tools belong here when they provide durable layers for continuity in agent behaviour, including conversation, entity, user, episodic, and workflow state.
Include when: The system models memory or evolving state for agents rather than only storing documents.
Exclude when: It is a generic database, cache, or vector store without an agent memory abstraction.
3 published tools
Payments
Payment infrastructure belongs here when it lets agents discover prices, hold or use payment credentials, transact, or operate within programmable financial controls.
Include when: An agent can initiate, receive, or negotiate a payment through a documented machine interface.
Exclude when: It is a consumer wallet, payment processor, or fintech API with no substantive agent role.
4 published tools
Frameworks & Standards
Frameworks and standards belong here when agents, agent runtimes, or agent interactions are their central subject.
Include when: The framework provides first-class agent abstractions, execution, tools, state, or coordination.
Exclude when: It is a general AI, machine-learning, or application framework with no first-class agent model.
4 published tools
Orchestrators
Orchestrators belong here when their main purpose is coordinating agent work across multiple runs, roles, workers, or workflows rather than implementing one agent in isolation.
Include when: The system assigns, schedules, routes, or supervises work across agents or agent runs.
Exclude when: It is a single-agent framework with no material coordination layer.
1 published tool
API Orchestration
API orchestration tools belong here when they sit between an agent and external APIs, and discovery, credentials, policy, routing, or execution materially improves safe tool use.
Include when: The product gives agents a documented way to discover and invoke APIs, tools, or skills.
Exclude when: It is a conventional API gateway or client with no substantive agent workflow.
1 published tool
Voice & Multimodal
Voice and multimodal tools belong here when real-time speech or another non-text modality is central to the agent’s operation or user experience.
Include when: The product provides a documented runtime or interface for voice or multimodal agents.
Exclude when: It is only a generic speech, image, or video model API.
2 published tools
Search & Discovery
Search and discovery tools belong here when their indexed data, ranking, enrichment, or result structure materially improves an agent workflow.
Include when: The search corpus or enrichment is specialized and documented.
Exclude when: It is a broad consumer search engine with no substantive agent interface.
3 published tools
Marketing & SEO
Marketing and SEO tools belong here when they are designed for agent workflows, including demand research, search analysis, content planning, optimisation, and measurable follow-up actions.
Include when: An agent can research, analyse, recommend, or execute a substantive marketing or SEO task through a documented interface.
Exclude when: It is a conventional marketing suite with only a generic API or superficial agent integration.
Methodology
How this snapshot is produced
- Source. Approved category and tool records originate in the public content repository and are published into the site's live index.
- Unit. Each currently published tool record counts once, under exactly one category.
- Eligibility. A record must pass the directory's evidence-based inclusion test and receive one classification.
- Counts. Totals are calculated from the same live records used to render the directory. They change when content is added, reclassified, unpublished, or removed.
- Reproduction. For a stable citation, record the access date and the content-repository commit inspected rather than treating this live page as an immutable release.
Limitations
What the data does not establish
- The directory is curated and incomplete; counts are not total market size.
- Inclusion and classification do not measure quality, security, reliability, revenue, usage, or adoption.
- Most listing facts are documentation-reviewed unless a dated hands-on test is explicitly described.
- Pricing labels are broad and can become stale between reviews.
- Category size reflects editorial coverage as well as real activity in that part of the ecosystem.
- A live snapshot cannot by itself support trend claims without comparable historical releases.
Taxonomy reference
All category guidance written for the directory is public
The current reference defines 16 known category slugs. Category definitions describe editorial scope; they do not imply that every category already has sufficient published coverage.
Found a missing record or a count based on stale content?
Report a correction