For AI agents: a documentation index is available at /llms.txt
AXO · Callable
Get your product into ChatGPT, Claude and Gemini
WorkPods runs the Callable layer of AXO (agent experience optimization), which gets your brand listed where AI assistants connect to live software tools. You need an app in ChatGPT, Claude or Gemini only when buyers want live data or actions inside the chat; if they do not, fix your AI search citations and your website first.
| Directory | You | Rivals listed |
|---|---|---|
| MCP Registry | No | 3 of 5 |
| Claude connectors | No | 2 of 5 |
| ChatGPT apps | No | 2 of 5 |
| Google connected apps | No | 1 of 5 |
# scoped remote MCP server
auth: OAuth 2.1
tool: check_availability
readOnlyHint: true
destructiveHint: false01
Decide when to build a connected app or wait
A ChatGPT app, Claude connector or MCP server (an open protocol with an official registry) is worth building when users want to take live actions in a chat. If your buyers mostly research vendors before they call sales, put GEO and agent-ready websites ahead of a custom app.
The guide below sets out where we think the Callable layer fits today:
| Business and buying motion | Should you build an app or connector now? | Why this is the right call for marketing and product |
|---|---|---|
| SaaS or service with live data buyers query (inventory, rates, quotes, account lookups) | Yes, worth scoping now | Buyers and users can get a useful answer from your own data inside ChatGPT, Claude or Gemini, without leaving the chat |
| Transactional or booking business (appointments, demos, reservations, order checks) | Yes, worth scoping now | An assistant can check live availability and trigger a safe booking or lookup step without forcing the user onto a separate web form |
| Complex B2B sale where buyers research and compare before calling sales | Not yet; wait on a custom app | Buyers want clear pricing, comparisons and third-party proof first, so GEO citations and Readable and Operable site basics come first |
| Early-stage product with low category awareness in search and AI answers | Not yet; build the base first | In our view, an assistant is unlikely to reach for the tool of a brand it barely knows from search or from your product pages |
02
Get listed where agents look
Assistants find callable tools in four directories: the official MCP Registry, the Claude connectors directory, the ChatGPT apps directory and Google's connected apps. A clear listing there tells an assistant what your software can do and when to call it.
Say a buyer asks an assistant to check live plan availability or pull a quote from several vendors. If a rival has a connector in those directories and your brand is missing from all four, the assistant can finish the task with the rival alone.
03
Improve your tool descriptions and scope your server
If you already have an MCP server or app, we review its tool descriptions and safety hints; if you have none, we scope one remote MCP server with OAuth 2.1. Either way, your engineers get a written specification, and assistants get clearer instructions on how to use your tools safely.
Safety hints are MCP annotations such as readOnlyHint (marking a tool that only reads data) and destructiveHint (warning before an action changes records). OAuth 2.1 is the standard way users grant an app access to their account. An operator edits every specification before your team sees it, and your team approves it in our approval log, as set out on how WorkPods works.
04
How we measure it
We check where you are listed against up to five rivals, and we review your tool specifications before and after each fix. Each result has a defined status.
We do not invent a score for how often assistants choose you.
- For Callable platforms, we check the MCP Registry, Claude connectors, ChatGPT apps and Google connected apps. We mark your brand and up to five rivals as Listed or Not listed in each directory.
- Each AXO check on your tool descriptions and safety hints is scored Pass, Partial, Fail or Not applicable. A missing optional file or connector is Not applicable, never a fail, and only a present, broken item gets Fail.
We never guarantee rankings, citations, app picks or agent traffic, and we never claim to measure how often ChatGPT, Claude or Gemini pick a brand's app. Closed platforms do not publish that data, and WorkPods is not a partner of OpenAI, Anthropic or Google. The free GEO and AXO audit includes a directory check for you and your rivals, delivered within 3 working days.
Questions
Questions buyers ask
Do we need an app in ChatGPT or a Claude connector right now?
Only if buyers or users want live data or actions inside their AI chat. Typical cases are checking availability, producing quotes, booking appointments or looking up an account. If your buyers mostly read comparisons and then call sales, fix GEO and your website first, and come back to an app once those basics pass.
Where do AI assistants look to find connected brand apps?
They look in four places: the official MCP Registry, the Claude connectors directory, the ChatGPT apps directory and Google's connected apps. MCP, the Model Context Protocol, is an open protocol for connecting AI to software, with its own official registry. We mark your brand and up to five rivals Listed or Not listed in all four.
What are MCP tool descriptions and safety hints?
Tool descriptions are plain-language notes on your connector that tell an AI assistant what each action does and what inputs it needs. Safety hints are MCP annotations, such as readOnlyHint and destructiveHint, that say whether a tool only reads data or changes records. We score each of these checks Pass, Partial, Fail or Not applicable.
Can you measure how often ChatGPT or Claude picks our app?
No. We never claim to measure how often ChatGPT, Claude or Gemini pick your app in user conversations, because those platforms do not publish that data to agencies or tools. You can count the requests that reach your own server. What we check is your Listed or Not listed status and your tool descriptions, before and after each fix.
What does WorkPods deliver if we do not have an MCP server yet?
If you have no connector and the decision guide says you need one, we scope one remote MCP server with OAuth 2.1 sign-in. We also write the tool descriptions and safety annotations, and an operator reviews them, so your engineers start from a checked specification. Nothing ships until your team approves it in our approval log.
Are you an official partner of OpenAI, Anthropic or Google?
No. WorkPods is not a partner of OpenAI, Anthropic, Google or any other platform owner. We work from the public protocols, the official directories and our own tests across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews and AI Mode. With nothing to sell on any platform's behalf, we can tell you plainly when to wait.