af agentic-first

Tony's agentics tool directory

Agentic tools.

A working directory of tools agents can actually operate. Good entries have a human view and an agentic surface, such as an API, MCP server, CLI, SDK, webhook, configuration path, or documented automation route that lets agents configure, update, maintain, observe, or control the system.

Agent access

Use the directory without scraping the page.

Agents can connect to the MCP endpoint for search and inspection, fetch the structured catalog directly, or read the compact machine index.

Data location

Choose tools with residency in mind.

Residency labels are conservative field notes, not legal advice. A country or region chip only appears when the exact product surface has an official region, endpoint, workspace, or deployment control. Adjacent product controls do not count; Verify vendor means the exact agent path still needs confirmation.

EU UK US Local Self-host Verify

Showing all 79 tools.

Protocols and standards

How agents connect and understand the rules.

Start here for common instruction files, tool protocols, site maps, user-interface streams, and agent-to-agent interoperability.

8 tools
ToolWhere to get itData locationUseful for
Model Context Protocol MCPprotocol Website
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Connects AI applications and agents to external systems, including tools, data sources, prompts, and workflows.
Agent Client Protocol Registry ACP Registryprotocol registry Registry
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Distributes ACP-compatible agents and adapters to clients that speak the Agent Client Protocol.
AGENTS.mdinstruction standard Standard
Customer controlledControl

Stored wherever the repository or website owner hosts it.

Provides repository-scoped instructions that coding agents can read before acting.
Agent2Agent A2Aagent communication protocol GitHub
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Enables independent agent systems to discover capabilities, communicate, and collaborate.
llms.txtllm readable site standard Standard
Site controlledControl

Public file; location follows the site hosting choice.

Publishes a curated, LLM-readable map of a site's key content so agents can find the right docs, references, and context.
The Website Specification Specification.websiteagent ready website standard Website
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Gives agents a platform-agnostic website quality spec covering agent readiness, AI discoverability, /llms.txt, Markdown pages, MCP access, accessibility, security, performance, privacy, and resilience.
AG-UI Agent-User Interaction Protocolagent ui protocol Website
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Connects agent backends to user-facing applications through standard event streams for chat, state sync, and human-in-the-loop UI.
AGNTCY Internet of Agentsmulti agent infrastructure standard Website
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Provides open infrastructure patterns for agent discovery, identity, messaging, and observability across vendors.

Agent communication

Where agents get inboxes, identities, and channels.

Email, messaging, agent-only chat, and collaboration layers for work that leaves the local machine.

10 tools
ToolWhere to get itData locationUseful for
AgentMailagent email infrastructure Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Provides API-first inboxes, threads, attachments, realtime events, skills, and MCP tools for AI agents that need to communicate over email.
Hostinger Agentic Mailagent email infrastructure Website
EU/US choiceEUUSGlobal

Hostinger hosting has EU and US regions; confirm email-specific storage before relying on it.

Provides hosted email infrastructure for agents and automation workflows, including webhooks, isolated inboxes, sender controls, and planned API/MCP connectivity.

Use isolated inboxes, allow/block lists, domain-level controls, webhook verification, and careful handling of credentials, OTPs, and customer mail. The validation report notes that full API/MCP features...

OpenMailagent email infrastructure Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Gives agents dedicated email inboxes with instant inbound handling, sender rules, attachments, CLI setup, and autonomous channel modes.
mails0agent email infrastructure Website
Self-host optionSelf-hostControlVerify

Self-hosting gives region control; hosted mailbox residency still needs vendor confirmation.

Provides hosted or self-hosted agent mailboxes with CLI, Python SDK, MCP server, agent skill docs, scoped keys, and verification-code extraction.
InboxAPIagent email infrastructure Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Gives MCP-compatible coding agents their own personal email address through a local CLI bridge and auto-created account.
A1Baseagent identity communications Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Gives agents real-world communication capabilities such as phone numbers, email, messaging platform participation, and shared chat history.
Relaycastagent messaging infrastructure Website
Local optionLocalControlVerify

Local daemon can keep state under operator control; hosted mode needs confirmation.

Provides shared channels, threads, DMs, reactions, files, search, inboxes, realtime events, webhooks, slash commands, and MCP tools for multi-agent systems.
AgentDMagent direct messaging Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Gives agents @aliases for direct and channel messaging over MCP and A2A, with Slack channel integration and tools for send, read, and delivery status.
Rineagent messaging network Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Provides durable, typed, signed, and encrypted messaging for agents with self-onboarding, proof-of-work registration, agent cards, WebFinger-style handles, and webhooks.
AgentMeshagent to agent mesh Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Lets agents register capabilities, discover peers, send requests, respond, emit events, and subscribe to updates through a common mesh protocol.

Coding agents and runtimes

Where the work actually runs.

Agent environments for inspecting projects, editing files, running commands, and moving from prompt to working code.

14 tools
ToolWhere to get itData locationUseful for
OpenAI Codexcoding agent runtime Website
Codex no DRUSGlobalVerify

OpenAI's ChatGPT residency article currently lists Codex as not supported for data or inference residency, so do not treat Codex as EU/UK-resident even if adjacent OpenAI products have regional controls.

Software engineering agent with CLI, web, automation, MCP, skills, hooks, subagents, and GitHub-oriented workflows.
Claude Codecoding agent runtime Website
US/globalUSGlobal

Claude Code is built on Anthropic infrastructure; first-party Anthropic workspace geo is currently US-only and inference controls are US or global, not EU/UK.

Agentic coding tool with MCP, subagents, hooks, worktrees, scheduled prompts, plugins, and external tool integrations.
OpenClawlocal agent orchestrator Docs
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Local-first agent system that can bind Codex and run external coding harnesses through ACP.
Gooselocal agent framework Docs
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Local, extensible AI agent for engineering workflows with MCP-based extension.
Kiroagentic development environment Website
EU/US + Web USEUUS

Kiro IDE enterprise controls can keep GA model traffic in US/EU geographies; Kiro Web preview stores content in US East.

Agentic IDE, CLI, and web workflow with specs, steering files, hooks, smart context, and MCP support.
Codeboltagent engine Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Agent engine for custom agents, tool calls, memory, plugins, guardrails, and multi-environment execution.
Cursoragentic code editor Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Reads codebases, edits files across projects, and runs agentic coding workflows inside an editor.
Clinecoding agent runtime GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Plans changes, edits files, runs terminal commands, and can drive browser workflows inside VS Code with approval steps.
Aidercoding agent runtime GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Pair-programs with LLMs in a local git repo, edits files, and can commit changes as work progresses.
Gemini CLIcoding agent runtime GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Inspects projects, edits files, and runs commands as an AI agent directly in the terminal.
Windsurfagentic development environment Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Reads repositories, plans multi-file edits, and verifies changes inside an agentic IDE workflow.
Ampcoding agent runtime Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Plans and executes multi-step coding tasks, edits repositories, and runs tests through CLI or VS Code workflows.
OpenCodecoding agent runtime Website
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Inspects projects, edits files, and runs commands as a provider-agnostic coding agent in the terminal.
Zedagentic code editor Website
Local defaultLocalVerify

Editor is local; hosted AI/collaboration features need residency confirmation.

Provides fast code editing with native agent workflows, external agent support, and MCP server connections.

Hosted AI and model platforms

Where agents call models, inference, and hosted runtimes.

Large language model APIs, inference clouds, model marketplaces, gateways, and regional app/GPU hosts that agents can call directly.

14 tools
ToolWhere to get itData locationUseful for
OpenAI Platform OpenAI APIhosted llm api platform API
Eligible regionEUUKUSVerify

OpenAI API data residency is a per-project control for eligible organizations. US and Europe support storage and processing; UK is storage-only in the current docs. Non-US regions require additional eligibility controls.

Hosted frontier-model API for agents that need chat, responses, tool calling, embeddings, multimodal generation, realtime voice, files, evals, and project-level data controls.
Anthropic Claude API Claude APIhosted llm api platform API
US/globalUSGlobal

First-party Claude API workspace geo is currently US-only. Supported newer models can route inference as US-only or global; use Bedrock, Vertex, or Foundry when those provider regions are the control surface.

Hosted Claude API for agents that need long-context reasoning, tool use, files, code execution features, managed agents, batch work, and workspace-level policy controls.
Google Gemini API / Vertex AI Gemini Vertexhosted llm api platform Docs
Cloud regionEUUKUSGlobal

Google documents selected-location storage and regional or multi-regional ML processing for listed Gemini Enterprise Agent Platform capabilities; unsupported endpoints do not guarantee in-region processing.

Google Cloud model and agent platform for agents that need Gemini models, regional endpoints, tool use, multimodal generation, embeddings, RAG, evaluation, and enterprise cloud controls.
Azure AI Foundry Models Azure Foundryhosted llm api platform Docs
Azure geoEUUSGlobal

Foundry deployment types distinguish global, data-zone, and regional processing; data at rest remains in the designated Azure geography for supported deployments.

Microsoft Azure model platform for agents that need Azure OpenAI, Foundry Models, regional deployments, Agents, enterprise IAM, networking, and compliance controls.
Amazon Bedrock Bedrockhosted llm api platform Docs
AWS regionEUUSJapanAustraliaGlobal

Bedrock runs in AWS regions and supports geography-specific cross-region inference profiles such as US, EU, APAC, Japan, and Australia; model providers do not access Bedrock prompts or completions.

AWS managed foundation-model and agents platform for model invocation, knowledge bases, guardrails, Bedrock Agents, tool use, evaluations, and enterprise cloud controls.
Mistral AI La Plateforme Mistral APIhosted llm api platform Docs
EU defaultEUUS

Mistral says data is hosted in the EU by default, with an explicit US API endpoint option and some feature-dependent transfers to listed subprocessors.

European model platform for agents that need Mistral models, agents/conversations APIs, tool calling, document intelligence, coding models, and hosted or self-hosted deployment choices.
Cohere Platform Coherehosted llm api platform Docs
Private deploySelf-hostControlVerify

Cohere documents customer-managed private deployments where interactions remain in the customer's infrastructure; verify standard SaaS/API residency before sensitive use.

Enterprise model platform for agents that need Command, Embed, Rerank, Compass, tool-aware retrieval, private deployments, and secure model serving.
GroqCloud Groqhosted llm inference platform Docs
US storageUS

Groq states customer data retained by GroqCloud is in GCP buckets located in the United States; ZDR can reduce retention for eligible features.

Fast hosted inference platform for agents that need OpenAI-compatible chat, responses, audio, compound systems, built-in tools, and low-latency model calls.
Together AI Togetherhosted llm inference platform Docs
VPC/privateControlVerify

Together documents private networking and VPC-based deployments for data-residency requirements, but public docs do not confirm a simple serverless EU/UK/US region selector for standard inference.

AI-native cloud for agents that need serverless inference, dedicated inference, fine-tuning, GPU clusters, OpenAI-compatible APIs, speech, image, and multimodal model access.
Fireworks AI Fireworkshosted llm inference platform Docs
Region choiceEUUSGlobal

Fireworks documents deployment multi-regions including GLOBAL, US, EUROPE, and APAC, with single-region availability by request or capacity.

Hosted and dedicated inference platform for agents that need OpenAI-compatible APIs, low-latency model serving, on-demand deployments, fine-tuning, metrics, and regional placement.
Hugging Face Inference Providers HF Inferencehosted model marketplace Docs
Provider-specificVerify

Hugging Face documents US/EU storage regions for Team and Enterprise Hub assets and Inference Endpoints, but Inference Providers are provider-specific and should be verified for the selected provider and route.

Hosted model marketplace and API layer for agents that need access to many open and commercial inference providers through Hugging Face SDKs, Hub models, and endpoint routes.
Replicatehosted model api platform Docs
US hostingUS

Replicate's subprocessor list shows infrastructure and service subprocessors located in the United States; API prediction data is automatically removed after one hour by default.

Hosted model API for agents that need quick access to image, video, audio, vision, document, and language models without managing inference infrastructure.
OpenRouterhosted llm gateway Docs
Enterprise EUEUVerify

OpenRouter documents enterprise EU in-region routing by request via the EU base URL. Each downstream provider and model route still needs review, so do not assume default OpenRouter traffic is EU-resident.

Model gateway for agents that need one API across many LLM providers, routing controls, privacy guardrails, ZDR enforcement, spend controls, and provider fallback.
Fly.io GPUs and Machines Fly.iohosted ai app runtime Docs
Region choiceEUUKUSControlGlobal

Fly.io lists selectable regions including Amsterdam, Frankfurt, London, and several US regions; Fly Machines and Volumes are tied to the region where they are created.

Global app and GPU runtime for agents that need to host model-adjacent services, inference workers, tool servers, APIs, private networks, and regional microVMs close to users.

Control planes and orchestration

How humans coordinate multiple agents.

Dashboards, bridges, work boards, containers, and orchestrators for supervising sessions and keeping approval paths visible.

10 tools
ToolWhere to get itData locationUseful for
OpenACPagent session bridge GitHub
Self-hostSelf-hostControl

Self-hosted or local deployment lets the operator choose region.

Bridges chat and messaging surfaces to ACP-compatible coding-agent sessions with streaming, switching, and approval flows as the project matures.

High-privilege bridge. Keep bot tokens secret, use permission gates, avoid public exposure, and sandbox production workspaces. The validation report found the repository real but very immature; verify which...

clidecklocal multi agent dashboard GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Local dashboard for running and coordinating multiple CLI coding agents from one browser window.
Codex Orchestratoragent orchestration GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Claude Code plugin and CLI for spawning Codex agents in tmux sessions and collecting results.
Composio Agent Orchestratormulti agent orchestrator GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Runs multiple coding agents across isolated worktrees with dashboard supervision.
Vibe Kanbanmulti agent work board GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Dispatches board-tracked work across coding agents in isolated workspaces and lets humans review diffs from one UI.
Conductormulti agent worktree control plane Website
Local defaultLocalVerify

Mac/worktree workflow is local by default; verify any hosted/team sync feature.

Runs multiple coding agents in parallel on a Mac, each in its own worktree, with diff review before merge.
Sculptormulti agent container orchestrator GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Runs multiple coding agents in isolated containers and syncs selected agent work back into the local repo.
container-useagent container isolation GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Gives each coding agent an isolated container and git branch, with command history and session inspection.
Claude Squadmulti agent terminal manager GitHub
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Manages multiple terminal coding agents in isolated tmux sessions and git worktrees, including background runs.
here.nowagent web hosting Website
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Lets coding agents publish files, folders, prototypes, reports, dashboards, documents, games, and static sites to live URLs in seconds via API, CLI, or the here.now skill.

Anonymous sites are public and temporary by default. Review what the agent publishes before sharing URLs, avoid secrets or private data in uploaded files, use password protection for sensitive previews, and review the install script/skill before...

Monitoring and usage

How operators keep an eye on cost, sessions, analytics, and drift.

Usage, status, token, trace, session, web analytics, and cost visibility for long-running agentic work.

9 tools
ToolWhere to get itData locationUseful for
AgentsViewsession browser analytics Website
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Local-first browser for past AI coding sessions, search, analytics, and usage reports.
OpenUsageusage cost monitoring Website
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Local terminal dashboard for tracking spend, quotas, rate limits, models, sessions, MCP usage, and code stats across AI coding tools.
Token Trackertoken usage monitoring Website
Local + cloud opt-inLocalVerify

Local by default; optional cloud sync/leaderboard needs residency confirmation.

Local-first token dashboard for many coding-agent CLIs, focused on counts, model names, timestamps, and project attribution.

Prefer token-count-only collection. Review hook installation and optional cloud sync before enabling. The validation report notes auto-installed hooks and opt-in cloud leaderboard sync.

ccusagetoken usage monitoring Website
Local defaultLocalControl

Runs locally by default; model/provider calls may leave the machine.

Analyses local coding-agent CLI usage data for directional token usage, estimated cost, cache creation/read tokens, and daily, weekly, monthly, or per-session reports.

Especially useful for spotting reusable prompt, skill, command, and agent-definition patterns that improve cache writes and cache reads.

Langfusellm agent observability Website
EU/US/Japan/self-hostEUUSSelf-host

Langfuse Cloud offers EU, US, Japan, and HIPAA US regions; self-hosting gives full region control.

Traces, monitors, and evaluates LLM and agent applications end-to-end, including cost, token, latency, and quality signals.
Heliconellm agent observability Website
EU/US/self-hostEUUSSelf-host

Helicone documents EU and US cloud regions and self-hosting for customer-controlled residency.

Monitors LLM and agent requests for cost, latency, usage, errors, and gateway-level behaviour.
Arize Phoenixllm agent observability GitHub
Self-hostSelf-hostControl

Self-hosted or local deployment lets the operator choose region.

Traces and evaluates agent and LLM applications locally or self-hosted, surfacing slow, costly, or low-quality spans.
OpenLLMetryllm telemetry GitHub
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Captures LLM-specific telemetry such as token counts, latency, model versions, errors, and traces on top of OpenTelemetry.
Matomo MCP Serverweb analytics mcp Docs
EU/self-hostEUSelf-host

Matomo Cloud is EU-hosted; On-Premise lets the operator choose the country/location.

Lets Codex, Claude, ChatGPT, and other MCP-compatible agents query self-hosted Matomo web analytics for traffic trends, campaigns, referrers, pages, conversions, and visitor behaviour.

Use a dedicated read-only Matomo user or token, prefer aggregate/anonymised reports, avoid visitor-level data by default, and treat URLs, referrers, titles, campaigns, search terms, and event metadata as...

Payments and transactions

How agents ask for, approve, and settle value.

Emerging payment protocols for agent-initiated API access, purchases, and commercial workflows.

2 tools
ToolWhere to get itData locationUseful for
Agent Payments Protocol AP2agent payment protocol Website
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Authorises and verifies agent-initiated payments across merchants, agents, and payment providers.
x402agent payment protocol Website
Impl. choiceControl

Open standard; data location depends on the implementation or host.

Lets agents pay for APIs and web resources over HTTP without traditional account setup or API-key exchange.

E-commerce and agentic commerce

Where agents discover products, build carts, and touch checkout.

Big commerce and payment platforms with agent toolkits, MCP surfaces, catalog access, or credible agentic-commerce programs.

7 tools
ToolWhere to get itData locationUseful for
Stripe Agent Toolkit Stripe agentsagentic commerce payments Docs
Verify vendorVerify

Stripe documents transfer frameworks, but this pass did not verify a simple EU/US storage choice for agent tooling.

Lets agents and agent frameworks call Stripe APIs, create payment links, work with MCP, and build agentic commerce flows.
Shopify Storefront Catalog MCP Shopify MCPecommerce catalog mcp Docs
Europe + globalEUGlobal

Shopify stores certain merchant/customer personal data in Europe while operating global transfer controls.

Exposes a Shopify merchant catalog to agents so buyers can search and discover products from a single storefront.
PayPal Agent Toolkit PayPal agentsagentic commerce payments Docs
Verify vendorVerify

PayPal has regional privacy terms, but this pass did not verify a clear EU/US storage selector for the agent toolkit.

Wraps PayPal APIs for orders, invoices, subscriptions, shipments, transaction details, and disputes in agent-callable tools.
Visa Intelligent Commerce Visa MCPagentic commerce network Program
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Connects AI agents to Visa Intelligent Commerce and Visa Acceptance APIs through MCP and an acceptance agent toolkit.

Pilot and partner access may apply. Treat as high-trust payment infrastructure: verify eligibility, tokenization, identity, audit, and user consent flows.

Mastercard Agent Payagentic payment network News
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Mastercard initiative for agentic payments and commerce flows, including collaboration with Microsoft AI infrastructure.

This is payment-network infrastructure, not a drop-in open-source package. Verify availability, integration route, fraud controls, identity, and user-consent requirements.

Adyen Agentic Commerceagentic commerce payments Website
Europe primaryEUUSGlobal

Adyen says payments processed are stored in Europe and an isolated India environment; systems also run in several regions.

Positions Adyen payment infrastructure for agentic commerce, AI-driven payments, and Agent Payments Protocol ecosystem work.

Good strategic signal, but verify concrete APIs, regional availability, and production integration paths before listing as an installable agent tool.

BigCommerce AI-ready Developer Docsecommerce platform api Docs
Hosted: verifyVerify

Hosted service; no official residency claim verified in this pass.

Gives agents API documentation, llms.txt indexes, and commerce APIs for storefront, catalog, checkout, orders, customers, and integrations.

Not a dedicated agent product. Treat store API credentials as privileged access to catalog, orders, customers, checkout, and settings.

CRM and customer operations

Where agents work with customers, cases, and revenue systems.

Large CRM, support, and enterprise workflow platforms that expose agent builders, MCP servers, or customer-data automation surfaces.

5 tools
ToolWhere to get itData locationUseful for
Salesforce Agentforcecrm agent platform Docs
Region choiceEUUSGlobal

Salesforce Hyperforce provides local data storage/residency choices; confirm product and org region.

Builds, manages, deploys, and integrates agents over Salesforce CRM, Customer 360, Data Cloud, Flow, Apex, mobile SDKs, and enterprise workflows.
HubSpot MCP Servercrm mcp server Website
EU/US + moreEUUS

HubSpot accounts are hosted in selected regional data centers including US and EU Germany.

Lets MCP-compatible AI assistants securely interact with HubSpot CRM data through scoped access and natural-language workflows.
Zendesk AI Agentssupport agent platform Docs
EU/US + moreEUUKUS

Zendesk offers regional hosting for entitled customers, including US and EEA choices.

Extends Zendesk support automation with developer-customizable AI agents and API-key authenticated platform access.
Microsoft Dynamics 365 Agents Dynamics agentscrm erp agent platform Docs
Azure geo choiceEUUSGlobal

Dynamics 365 stores customer data at rest in the selected Azure geo, with EU Data Boundary commitments.

Uses Dynamics 365 and Copilot Studio agents for sales, service, finance, operations, contact center, and customer journeys.
ServiceNow AI Agent Studioenterprise workflow agent platform Website
Region choiceEUUSGlobal

ServiceNow has EU hosting options and AI data residency controls; confirm contract/product scope.

Creates, manages, tests, and observes AI agents and agentic workflows across ServiceNow business processes.

Agent access

Agents can query this directory directly.

The directory is already exposed as a Streamable HTTP MCP server, so agents do not need to scrape this page. Use it to search by group, category, data location, residency posture, deployment model, works-with tag, or open-source status.

Legend

What the labels mean.

Tony's broader framing: start with safe AI lanes, then the Agentic-first website playbook and the AI-readable data room essay.

Directory groups

Protocols
Standards and shared contracts agents use to discover, talk, pay, or understand a site.
Communication
Agent inboxes, identity, email, messaging, and collaboration channels.
Runtimes
Places where agents inspect projects, edit files, run commands, and do work.
Hosted AI
Model APIs, inference clouds, gateways, and hosted runtimes agents can call.
Control planes
Dashboards, bridges, schedulers, and orchestration layers for managing agents.
Monitoring
Usage, token, trace, analytics, session, cost, and observability tools.
Payments
Protocols for agent-initiated payment and value exchange.
Commerce
Catalog, checkout, order, and merchant platforms with agent-facing surfaces.
CRM
Customer, support, revenue, and workflow systems agents can operate through APIs or MCP.

Data location chips

EU
Official docs show EU/EEA control for the exact agent-facing product surface.
UK
Official docs show UK control for the exact surface, sometimes storage-only.
US
Official docs show US residency, US hosting, or a US-only/default surface.
Local
Runs on the operator's machine by default; connected models or sync may still leave.
Self-host
Can run on infrastructure the operator chooses.
Control
Location depends on the implementation, host, repository, or customer configuration.
Global
Global, multi-region, marketplace-routed, or not pinned to one jurisdiction by default.
Verify
No clear public claim was confirmed for the exact API, MCP, runtime, or workload path.

Exact-surface rule: do not infer residency from headquarters, CDN location, marketing copy, or an adjacent product. The chip must apply to the agent/API/MCP/runtime path you are actually using.

Feedback

What should be added next?

Got a tool that belongs here? Update me on X with what it is and how agents can configure, update, maintain, observe, or control it.