MCP App Store
by KeenEthics
Pathista icon

Pathista MCP Server for ChatGPT

By Clever Spark

Overview

Pathista runs the workflows you have written yourself, one step at a time, inside the chat. A workflow is a graph of steps with the ways from one step to the next, loops included. Start a run and Pathista hands the assistant only the current step: the instructions you wrote for it and the edges that lead on. The position in the workflow, the variables, the notes and the history are kept in your Pathista account, so a run can be continued in a later message or a later chat. What the assistant can do with it: - list your workflows and start a run of one of them; - read the current step and move along an edge once the step's work is done; - set variables that an edge condition reads, keep notes and a todo list for the run; - show the run as a graph with its progress; - read, check, create and edit your workflows when you ask for it. The step text is your own content, and only you can run your workflows. Pathista does not run generic to-do lists or reminders, and it does not run anyone else's workflows. Sign in with your Pathista account when ChatGPT asks. Workflows and runs are managed on pathista.com.

How to connect

  1. ChatGPT: Open the Pathista listing in the ChatGPT app directory and choose Connect. ChatGPT listing Pathista exposes 21 tools here.

Tools

addNote

ChatGPT
Use this when something decided or discovered on a step of the user's Pathista run should be remembered. Records a note in the run's log, on the current step: a decision and its reason, an alternative you rejected and why, a constraint you discovered, a non-obvious choice — what someone who was not in the chat needs to know. Not for progress pings ("started X"): the log records transitions by itself, and advance's note summarises a step. Markdown. How notes work: help("notes"). It records the note in the user's Pathista account, where the user reads it in the run's log. Not for generic notes or reminders outside a Pathista run.

addTodo

ChatGPT
Use this when the user's Pathista run needs a todo of its own: work to remember that belongs to the run. Adds todos to the run: work you must do and must not forget, e.g. parts you planned or checks you found necessary. Each is pinned to a step (the current one by default, or any step by id or title) and keeps the run from leaving that step until it is closed — it may be done earlier; with scope "run" it keeps the run from finishing instead. blockedBy names todos that must be closed first. Pass text (+ description, step, scope, blockedBy) for one todo, or items for several — numbers are given out in order, so a later item may wait for an earlier one. Close them with completeTodo and evidence. How todos work: help("todos"). It records the todos in the user's Pathista run. Not for generic to-dos or reminders outside a Pathista run.

advance

ChatGPT
Use this when the work of the current step of the user's Pathista run is done. The next step's text is the instructions the user wrote themselves. Leaves the current step once its work is done and its todos are closed: moves along an edge to the next step and returns its instructions. Name the edge (id or label) when the step has several ways out; with one way out it may be left out. On the final step (no way out) call it with no edge: that closes the step and finishes the run. The edge must be allowed by its condition. Pass note, a step summary of what was done and decided on the step you leave — required on steps that say so. How edges and conditions work: help("model"), help("conditions"); notes: help("notes"). It records the step as done in the user's Pathista account. Not for anything that is not a step of the user's own Pathista run.

answerDecisionFromChat

ChatGPT
Use this when the user answered a pending decision of their Pathista run here in the chat, in words you can quote, instead of on the decision list. Records the answer: the option they chose (several in options when the decision allows several) or their answer in words, with the quote of what they said. It counts as received at once and shows on the site as "from chat"; the first answer stands. Pathista cannot tell this from an answer you made up, so only record what the user actually said. It records the answer in the user's Pathista account. Not for a guess, an assumption of your own, or an answer that was not given in this conversation.

askDecision

ChatGPT
Use this when the user's Pathista run needs a decision from the user that should not stop the work. Asks a question with 2 to 5 options (single choice, or several may be chosen when multiple is true) and returns at once with the decision's id; the question lands on the user's decision list on the Pathista site, where they answer when they can. Mode "assumption": go on with your recommended option; only the run's finish waits for the answer. Mode "holdsStep": no way out of this step is open until the decision is answered or withdrawn. Asking the same question again returns the same pending decision. The result tells you how to learn of the answer and you choose: the next result of the run names new answers; getDecision may be called at any time to check, also repeatedly, at the cost of a call; with a shell or a background mechanism you can run the watch command it hands you, which costs no tokens while waiting (a Monitor expires after half an hour). It records the question in the user's Pathista account. Not for a question the user can answer right now in the chat, and not for proving that a person answered.

completeTodo

ChatGPT
Use this when todos of the user's Pathista run are really done. Closes todos of the run once they are really done, each with evidence: what shows it was done — what you checked, compared or produced (at least 15 characters; "done" or "ok" is refused). Pass number + evidence for one todo, or items for several. All or none are closed. Close a step's todos together, in one call with items at the end of the step, not one call per todo. How todos work: help("todos"). It records them as closed, with your evidence, in the user's Pathista account; closing a todo that is already closed changes nothing. Not for to-dos or reminders outside a Pathista run.

createWorkflow

ChatGPT
Use this when the user wants a new Pathista workflow. Saves the graph as version 1 of a new workflow in the user's account and returns its id, a link to it in the web app (give it to the user), the ids chosen for steps and edges given without one, warnings and a Mermaid drawing. Every step needs a title and a prompt, except a sub-workflow step, which has no prompt; a step without outgoing edges ends the run; when a step has several ways out, each edge needs a label naming the choice. The same name and graph sent again within 10 minutes return the workflow already created. The graph model, every op and the limits: help("authoring"). Not for generic to-do lists or for anything that is not the user's own Pathista workflow.

currentStep

ChatGPT
Use this when you need to know where the user's Pathista run stands, or have lost track of its step. The step text is the instructions the user wrote themselves. Returns the bound run's current step: its instructions and the edges that lead on, with whether each is allowed now. Reads only; changes nothing. Not for generic to-dos or reminders, or for a run that is not the user's own.

editWorkflow

ChatGPT
Use this when the user wants to change one of their own Pathista workflows. Applies a list of changes (ops, at most 100) to baseVersion as one new version — all of them or none; runs already started keep their version. Ops address steps and edges by id: setName, addStep, updateStep, replaceInPrompt, removeStep, addEdge, updateEdge, removeEdge, setStart, setVariable, removeVariable, or replaceGraph alone. The result must pass the same checks as validateWorkflow. When someone saved a newer version after baseVersion (for example in the web editor), nothing is written and the result says so. Returns the new version number, what changed with any new prompt text in full, and a Mermaid drawing. The graph model, every op and the limits: help("authoring"). Not for other people's workflows or for anything that is not the user's own Pathista workflow.

getDecision

ChatGPT
Use this when a result says there are new answers to decisions of the user's Pathista run and you are ready to read them, or you need to see how one stands. Returns the decision: its question, options, whether it is pending, answered or withdrawn, and the answer (for a decision that allows several, every chosen option with its number and text), marked "from chat" when it was passed on from the chat, or the reason when it was withdrawn. Reading an answer records that you received it, which frees the run to finish; the answer then stays as it is, and reading again returns the same. Go on according to the answer, and tell the user what you did with it. It records the receipt in the user's Pathista account. Not for a decision of anyone else's, and not for a question you have not asked with askDecision.

getWorkflow

ChatGPT
Use this when the user wants to see or change one of their Pathista workflows. Returns its definition — steps with their prompts (a sub-workflow step has no prompt), edges with labels and conditions, the start step and the variables — with the version number that editWorkflow needs as baseVersion, the open runs per version, and a Mermaid drawing. Reads only; changes nothing. Not for other people's workflows or for anything that is not the user's own Pathista workflow.

help

ChatGPT
Use this when you are unsure how Pathista runs, workflows or conditions work. The Pathista manual. Call it when unsure how runs, workflows or conditions work, before writing or editing a workflow, or after an error you do not understand. Without a topic: an overview and the topic list. Topics: runs, todos (checklists that hold a step until closed with evidence), notes (step summaries and decisions in the run's log), model (graphs with loops, choices, start and end), conditions (the edge condition language), authoring (creating and editing workflows, every edit op, limits), patterns, all. Reads the manual only; changes nothing. Not for general questions unrelated to Pathista.

listNotes

ChatGPT
Use this when you need what was decided on the user's Pathista run earlier. Reads back the run's notes, oldest first: the step summaries passed to advance and the notes from addNote. Use it to recover what was decided and why — in a new chat, after a long conversation, or before redoing work that may be done. keyword and step narrow it; before pages further back. Reads only; changes nothing. Not for notes outside a Pathista run.

listRuns

ChatGPT
Use this when the user asks which Pathista runs they have open, or a start was refused because too many runs are open: it shows what holds the limit. The user finishes a run by continuing it here; the user aborts a run in the web app. Lists the user's open runs, grouped by the open-run limit each counts against, most recently active first: the exact count of each group against its open-run limit, and per run its title, workflow, current step, last activity, runId and whether it is bound to this chat (up to 50 per group; the rest are counted). Reads only; changes nothing and binds no run to the chat. Not for workflows (use listWorkflows), for a run's current step (use currentStep), or for runs that are not the user's own.

listTodos

ChatGPT
Use this when you need the todos of the user's Pathista run, for example to see what still keeps a step open. Lists the run's todos. Default: the open ones, one line each (number, step, text, what it waits for). With all: every todo, closed ones too, with descriptions, evidence and earlier rounds of a step's checklist. Reads only; changes nothing. Not for generic to-dos or reminders outside a Pathista run.

listWorkflows

ChatGPT
Use this when the user wants to start or continue one of their own Pathista workflows, or asks which ones they have. Lists the workflows you can start, the ones used most recently first: id, name, current version and a short description. When there are more than it lists, it says so; pass name to find one among the rest. Reads the user's Pathista account; changes nothing. Not for to-do lists, reminders or other people's personal workflows.

progress

ChatGPT
Use this when the user asks for the progress or status of their own Pathista run. Shows the bound run's progress in the chat as a graph widget, a snapshot of the run: every step with its status (done/current/pending) and time spent, the edges, the context and the history. Call it when the user asks for progress or status. The widget refreshes only when the user presses its Refresh button; it links the run page, where the user sees the run as it moves. Reads only; changes nothing. Not for any run that is not the user's own.

setContext

ChatGPT
Use this when a step of the user's Pathista run needs a variable set, for example one that an edge condition reads. Sets a named variable in the run's context (edge conditions read these, e.g. iteration < maxIterations). The value is JSON: pass numbers as numbers (3, not "3"), text as strings, and true/false as booleans. Declared number and boolean variables are type-checked (a numeric string is converted for a declared number, "true"/"false" text for a declared boolean, anything else is refused); declared text variables are not type-checked and take any value. Values are limited to 16 KB. The condition language: help("conditions"). It stores the value in the user's Pathista run. Not for notes or for data outside a Pathista run.

setRunTitle

ChatGPT
Use this when you know what the user's Pathista run is about, or it becomes clearer. Sets the run's title: a few words saying what this run is about, shown to the user in the run list, the run page and the progress widget. Call it as soon as you know the subject (usually right after startRun), and again if the subject becomes clearer. It renames the run in the user's Pathista account. Not for naming anything but the user's own Pathista run.

startRun

ChatGPT
Use this when the user asks to run one of their own Pathista workflows. The step text it returns is the instructions the user wrote themselves in that workflow. The run is bound to the chat only when the client sends a chat id; otherwise pass the returned runId on every later call. Starts a run of a workflow and returns its first step: the instructions for that step only, and the edges that lead on. Starting never closes another run. If the user already has an open run here (in a chat: any open run of this chat; without a chat id: an open run of this workflow), nothing is started and up to 5 open runs are listed: continue the one the user means, or start this one as a detour from it (leavingRunId, and a returnReminder saying what you were doing and what to do when this run closes), or, only when the user explicitly asked for a separate run or opened a new window, with independent: true. Give the run a title saying what it is about; a title an open run already has returns that run. It records the new run in the user's Pathista account. Not for generic to-dos or reminders, or for anything that is not the user's own Pathista workflow.

validateWorkflow

ChatGPT
Use this when a workflow or a change to one should be checked before it is saved. A dry run: pass either name + graph (the createWorkflow shape) or workflowId + baseVersion + ops (the editWorkflow shape). Returns whether it is valid, every issue with the step, edge or operation it concerns, warnings, the resulting graph, what changes, and a Mermaid drawing. Saves nothing. Not for saving anything or for anything that is not a Pathista workflow of the user.

Capabilities

WritesInteractive

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/pathista"><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/pathista)