MCP App Store
by KeenEthics
Hamster icon

Hamster MCP Server for ChatGPT

By Hamster

Overview

Use Hamster to write PRDs, specs, and briefs as a team, then plan and ship them.

How to connect

  1. ChatGPT: Open the Hamster listing in the ChatGPT app directory and choose Connect. ChatGPT listing Hamster exposes 42 tools here.

Tools

archive_initiative

ChatGPT
Archive an initiative (soft delete — sets archived_at, matching the web app's archive action). Archived initiatives disappear from list_initiatives. Initiatives cannot be hard-deleted.

ask_hamster

ChatGPT
Use this when the user wants to ask Hamster something or have Hamster help with a brief. Creates a Hamster thread with prompt as the initial message, or adds prompt to an existing thread_id, and returns Hamster's reply. Pass brief_id to set the brief context. If outcome is pending, wait retry_after_ms, then call get_hamster_reply with the returned thread_id and turn_id. Each call adds a message; do not resend to check status.

create_brief

ChatGPT
Create a new brief with its backing document. Optionally pass markdown content to populate the brief's editable document body, and an initiative_id to link the brief to an initiative on creation. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

create_initiative

ChatGPT
Create a new initiative with its backing document. Optionally pass markdown content to populate the document body. Requires the initiatives.manage permission. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

create_note

ChatGPT
Create a document-only note. Optionally pass a description and markdown content to populate its editable body.

create_subtask

ChatGPT
Create a subtask under a parent task. Auto-increments parent's total_subtasks count. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

create_task

ChatGPT
Create a new top-level task. Display ID and notes document are auto-generated. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

delete_subtask

ChatGPT
Delete a subtask. Auto-decrements parent's total_subtasks (and completed_subtasks if it was done). If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

explore_entity

ChatGPT
Explore all relationships for a named entity in the knowledge graph. Pass an entity name or title and receive the matching context nodes plus their typed edges with direction, lane, and confidence. Use this when you know the NAME of an entity (e.g. "Q3 OKR Plan", "Customer Onboarding Initiative") and want to see everything connected to it — outgoing and incoming relationships, edge types, confidence scores, and lanes (asserted, derived, inferred). For topic search across the knowledge graph, use search_knowledge_graph instead. For raw node-ID traversal when you already have context node IDs, use get_neighborhood.

generate_plan

ChatGPT
Start plan generation for a brief through the same path as the web app's Generate plan button (requires documents.manage on the account). Once generation starts, it cancels any queued or running delivery of the brief. Returns the plan thread_id immediately; generation runs in the background, so poll get_plan with the brief_id until plan_generation_status is completed or completed_with_gaps (finished; the progress message names parent tasks left without subtasks) and read the tasks from there. Fails with a conflict while a generation is already running. To also deliver once the plan lands, use trigger_delivery instead — it generates the plan itself when needed. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

get_brief

ChatGPT
Get a brief by ID with its full document details. Pass include_content: true to also return the document body as markdown.

get_document

ChatGPT
Get a document by ID with all metadata. Pass include_content: true to also return the document body as markdown (works for any document type: brief, blueprint, method, note, ...).

get_goal

ChatGPT
Get a goal by ID. Pass include_content: true to also return the goal's document body as markdown.

get_hamster_reply

ChatGPT
Fetch the reply for a turn that ask_hamster reported pending; not needed otherwise. Pass exactly the returned thread_id and turn_id. If outcome is pending, wait retry_after_ms and repeat this read-only call with the same arguments. For failed or interrupted outcomes, response_complete is false and message explains whether returned text may be incomplete.

get_initiative

ChatGPT
Get an initiative by ID. Pass include_content: true to also return the document body as markdown.

get_neighborhood

ChatGPT
Resolve affinity edges between a set of seed context node IDs and the rest of the knowledge graph. Returns flat edge rows with lane and confidence so you can use them as graph priors. Use this AFTER you already hold seed node IDs (for example from explore_entity nodes or search_knowledge_graph evidence) to enumerate linked documents — including the inferred lane's "should be linked but isn't yet" candidates. Seeds must be context node UUIDs (at most 50); lanes must be asserted, derived, or inferred. Do not use this for absence checks or generic search. For topic search use search_knowledge_graph; for a named entity use explore_entity.

get_next_task

ChatGPT
Get the next actionable task. Filter by brief, assignee, statuses (default: [todo]), excluded_statuses, exclude_blocked (default: true), or assigned_to_me (auto-resolves the authenticated user). Custom priority ordering supported. Response includes assignee_id.

get_note

ChatGPT
Get a note by its document ID. Pass include_content: true to also return the note body as markdown.

get_plan

ChatGPT
Get a brief's plan: the brief status (including plan_generation_status) plus all generated tasks. Start generation with generate_plan or trigger_delivery, then poll this tool until plan_generation_status is completed or completed_with_gaps; both are finished, deliverable plans, and completed_with_gaps names the parent tasks left without subtasks in the progress message.

get_task

ChatGPT
Get a task by UUID or display ID (e.g. HAM-123). Returns the task with its subtasks and dependencies.

list_accounts

ChatGPT
List all team accounts you have access to. Does not mark which one is active; call switch_account to choose it.

list_briefs

ChatGPT
List briefs for the active account with document info (title, description, status, owner_id, plan generation status). Prefer filters (search, owner_id, assigned_to_me, statuses, excluded_statuses, tags, updated_after) over unfiltered calls — results are paginated (default 50 per page, max 200; page via offset until has_more is false), so an unfiltered call on a large account only returns the most recently updated page.

list_documents

ChatGPT
List documents. If parent_document_id is given, lists children. Otherwise lists root documents, optionally filtered by type. Prefer filters (document_type, search, owner_id, assigned_to_me) over unfiltered calls — results are paginated (default 50 per page, max 200; page via offset until has_more is false). Brief-backed rows include brief_id for get_brief navigation. document_type: "note" results include archived notes; only list_notes filters them out.

list_goals

ChatGPT
List non-archived goals for the active account (id, title, status, target date, parent/owner ids). Prefer filters (search, statuses, excluded_statuses) over unfiltered calls — results are paginated (default 50 per page, max 200; page via offset until has_more is false).

list_initiatives

ChatGPT
List initiatives for the active account (id, title, status, health, owner_id). Use the id as initiative_id when linking a brief via create_brief or update_brief. Default omits archived rows (Archive action, state=archived, or legacy status=archived). Pass statuses=["archived"] to list those. Prefer filters (search, statuses, excluded_statuses, health) over unfiltered calls — results are paginated (default 50 per page, max 200; page via offset until has_more is false).

list_members

ChatGPT
List members of the active account (user_id, name, email, role). Use user_id as owner_id when assigning an owner via create_brief, update_brief, create_initiative, or update_initiative. Results are paginated (default 50 per page, max 200; page via offset until has_more is false).

list_notes

ChatGPT
List notes for the active account. Prefer filters (search, owner_id, assigned_to_me) over unfiltered calls — results are paginated (default 50 per page, max 200; page via offset until has_more is false). owner_id and assigned_to_me filter by assignee, not creator. Archived notes are excluded by default; pass archived: true for archived only, include_archived: true for both (include_archived wins when both are set).

list_subtasks

ChatGPT
List subtasks for a parent task, ordered by position.

list_tasks

ChatGPT
List top-level tasks for the active account. Excludes subtasks. Filter by brief, statuses, excluded_statuses, priorities, assignees, exclude_blocked, or assigned_to_me (auto-resolves the authenticated user). Custom priority ordering supported. limit is capped at 500 per call; paginate with offset for more.

search_knowledge_graph

ChatGPT
Structured topic search across this account's knowledge graph (briefs, documents, threads, initiatives, and related evidence). Use this when you need retrieved evidence with relevance, confidence, and relationship context — typed hits you can cite, filter, or traverse. Each evidence item carries sourceRef.nodeId, which seeds get_neighborhood. Pass source_type to restrict document/thread kinds, brief_id to search only sources attached to one brief, and limit (1-20, default 5) to cap both direct hits and related entities. For multi-hop planning, or for open-ended questions or multi-step requests that need Hamster's judgment across the workspace's briefs, plans, tasks and knowledge, use ask_hamster (or your own reasoning over this tool's structured results). For a named entity's relationships, use explore_entity.

switch_account

ChatGPT
Switch the active account context. All subsequent tool calls (including later MCP requests from any client connected as you) will use this account until switched again.

trigger_delivery

ChatGPT
Deliver a brief: runs the web app's deliver-brief flow (spec gate -> plan generation if needed -> dispatch) with documents.manage on the account. All overrides are optional and fall back to the account's delivery settings. delivery_mode: auto runs the full chain; approval generates the plan and stops for a human to approve in the web app. method: cloud dispatches to a Hamster cloud sandbox that implements the plan, pushes a branch to the repository, and opens a pull request unless create_pull_request is false; local returns instructions for running the delivery on the caller's machine via the Hamster CLI plugin (/hamster:ship); an explicit method always wins over the account default. If the response has status method_choice_needed the account has no usable default — re-invoke with method set. Cloud deliveries return the brief thread_id and execution_id; follow plan generation with get_plan. Never call this repeatedly to poll: a second call while a delivery is in flight is rejected by the flow. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

update_brief

ChatGPT
Update a brief's status, owner, title, or description. Optionally pass markdown content to replace the persisted document body (omit to leave unchanged; empty string clears). An open editor converges to that body once it receives the update; unsaved in-editor keystrokes that were never stored may still merge back in. Pass an initiative_id to link (or re-link) the brief to an initiative. Steps apply sequentially (link, brief fields, document fields, content) — on error, earlier steps may already be persisted. Other fields are set-only; empty values are ignored, not cleared. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

update_initiative

ChatGPT
Update an initiative's title, description, status, state, health, focus, owner, or dates. Optionally pass markdown content to replace the persisted document body (omit to leave unchanged; empty string clears). An open editor converges to that body once it receives the update; unsaved in-editor keystrokes that were never stored may still merge back in. The field patch commits atomically before the content seed — on seed error the patch is already persisted. Other fields are set-only; empty values are ignored, not cleared. Requires the initiatives.manage permission. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

update_note

ChatGPT
Update a note's title, description, or archived state, and optionally replace its markdown body. Fields are set-only; omitted fields leave the current value unchanged, and empty title or description strings are ignored, not cleared. Pass archived: true to archive, false to unarchive. content replaces the persisted note body (omit to leave unchanged; empty string clears). An open editor converges to that body once it receives the update; unsaved in-editor keystrokes that were never stored may still merge back in. Steps apply sequentially (document fields, then content) — on error, earlier steps may already be persisted.

update_subtask

ChatGPT
Update a subtask's fields including status. If status changes to done, parent's completed_subtasks auto-increments. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

update_task

ChatGPT
Update task fields (title, description, priority, assignee, complexity, estimated hours, due date, execution type, executor). Does not change status — use update_task_status for that. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

update_task_status

ChatGPT
Change a task's status (todo, in_progress, done). Uses atomic status + reorder logic. If the workspace is connected to an issue tracker such as Linear, Jira, or Asana, the change can sync there too.

Capabilities

Writes

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/hamster"><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/hamster)