MCP App Store
by KeenEthics
Waypoint icon

Waypoint MCP Server for ChatGPT

By Silent Machines, Inc.

Overview

Waypoint lets your AI do real work. Ask like you normally do, and Waypoint picks the right tool from hundreds in its catalog: recent research papers and who wrote them, regulations and government notices, court records, company and market data, economic statistics, the weather, web pages and more. You never sign up for any of them. Results come back as readable tables and cards you can keep working with. Waypoint remembers along the way. Everything your tools find is kept in your Waypoint Library, so you can find it again in a later chat without fetching it again. It also remembers what you ask it to, such as your role or how you like answers written. You can see and delete all of it at waypoint.ing. Browsing the catalog works without an account. To run a lookup, connect your Waypoint account; you can create one in the same step.

How to connect

  1. ChatGPT: Open the Waypoint listing in the ChatGPT app directory and choose Connect. ChatGPT listing Waypoint exposes 9 tools here.

Tools

browse_catalog

ChatGPT
Show the user the full Waypoint catalog as an interactive card they can browse and pick from. Use this when the user wants to see what Waypoint can do, or asks you to show them the catalog or their options, optionally prioritised to one persona area via 'persona_group'. This is for a deliberate browse, not for searching mid-task: to find a specific capability while working on a task, use search_capabilities, which returns results without showing a card. Once the user picks a capability, call describe_capability for its exact price and input schema, then invoke_capability to run it.

describe_capability

ChatGPT
Get the exact input schema and credit price for a capability. It comes after search_capabilities and before invoke_capability, which takes its input in the schema returned here. Returns the capability's JSON Schema and the price as a range in Waypoint credits: a call reserves the top of the range and refunds the difference once the call reports what it actually used. This tool is free, spends nothing, and needs no account, so a price can be quoted before the user has signed up. When the user is signed in it also returns their balance as credits_remaining. Every capability has a kind, and some carry guidance: a short note on how to build good input for it. Waypoint's own kinds take an optional 'input' and answer with a plan: for kind quote, pass {steps: [{agent, capability}, ...]} to get one table pricing every step, with the total, how many steps the daily free allowance may cover, and, when signed in, the balance left after the worst case. For kind capability and kind report, 'input' is ignored; report runs with invoke_capability.

get_agent

ChatGPT
Fetch the details of a single Waypoint-registered agent by its URI. Use this after search_capabilities returns a promising row, passing that row's 'agent' value, or when the user explicitly names an agent (e.g. 'tell me more about acme/translator'). Returns the agent's record (id, owner, status, description, version, and compliance details where declared) and the capabilities Waypoint sells from it, each with its description, price and free-tier eligibility. For an agent Waypoint does not sell from, it returns the tools the agent exposes when Waypoint has probed them. Arguments: pass either 'agent_uri' (e.g. 'agent://acme/translator' or the short 'acme/translator' form) or the separate 'owner' + 'name' pair. Returns a single agent object, not a list. To run one of this agent's capabilities, call describe_capability for its price and schema, then invoke_capability.

get_balance

ChatGPT
Check the user's Waypoint credit balance. Use this when the user asks how many credits they have, or before a workflow that will make several paid calls. Returns the current balance in Waypoint credits. This tool is free and spends nothing. Credits are not sold through this tool or this connector: the user tops up at waypoint.ing.

get_result

ChatGPT
Fetch the outcome of a capability call that returned a job handle instead of a result. Takes the 'handle' from invoke_capability; a running job is usually polled a few seconds apart. Returns the result once the call has finished, or a status saying it is still running. Polling is free and spends nothing: the credits were already reserved when the call started, and the final charge is reported here.

invoke_capability

ChatGPT
Run a capability from the Waypoint catalog. Each call spends the user's Waypoint credits. describe_capability returns a capability's price and the input schema this tool takes. Capabilities are data lookups: each one reads from a data source and returns what it found. A call reserves the top of the quoted price and refunds the difference once the capability reports what it actually used; a failed call is refunded in full. Returns the result directly when it is ready quickly, otherwise a job handle for get_result. 'request_id' is an idempotency key: a new value for each distinct call, and the same value when retrying a call that timed out.

read_workspace

ChatGPT
Read the user's Waypoint workspace: the data they acquired through Waypoint and their history with it. Actions: records (search acquired data with query and filters), get_record (one record by id), history (recent Waypoint calls), runs (past tasks, or one task by id), recall (facts the user saved about themselves and past work relevant to query; with no query, their saved profile). Call recall at the start of a task that may build on earlier work. Before paying to fetch something again, check records: the user may already have it. Reads only the signed-in user's own workspace, changes nothing and spends no credits.

search_capabilities

ChatGPT
Search Waypoint to discover capabilities that can help with a task. Use this when the user asks about finding an agent, service, or API to solve a problem, or when you need to delegate a capability you do not have. Returns a ranked list of capabilities, each with the agent that serves it and, where filed, its category. One agent may serve many capabilities, so several rows can share an agent. Use 'intent' for natural-language queries (e.g. 'summarize research papers'); combine with 'capability' or 'tool_query' for stricter filtering. Call get_agent on a row's 'agent' value for its full provenance, tools and compliance detail. Call with no query to browse the whole catalog: it returns a rollup of every category and its capability count, plus the catalog itself. Pass 'category' to list or search just one category, 'agent' to list one agent's capabilities, or 'persona_group' to prioritise one area for the user. A query's 'total_capabilities' counts only the capabilities that query matched, not the whole catalog; call with no query to get the true total. To show the user an interactive catalog card they can browse and pick from, call browse_catalog instead; this tool returns results as data without rendering a card. Once you have picked a row, call describe_capability for its exact price and input schema, then invoke_capability to run it. Each row has a kind. Most are capability, one data call. Waypoint's own kinds do more: quote prices several capabilities in one table, and report turns the results a task fetched into one table.

write_workspace

ChatGPT
Change the user's Waypoint workspace. Actions: save_note (keep a note), delete_record (delete acquired data), forget (remove something Waypoint remembers). save_note keeps what the user asks you to: a note, or facts about them (facts, each with its type). delete_record can be restored for 30 days in Settings; forget cannot be undone. Confirm deletes and forgets with the user first. Spends no credits. 'request_id' must be a NEW unique string per distinct change, and the SAME string if you retry one that timed out.

Capabilities

InteractiveWrites

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/waypoint"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/waypoint)