PipesHub MCP Server
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| PIPESHUB_CLIENT_ID | No | OAuth app client ID (used with client secret for OAuth authentication) | |
| PIPESHUB_BEARER_TOKEN | No | JWT bearer token for authentication | |
| PIPESHUB_INSTANCE_URL | No | Your PipesHub instance URL, e.g. https://app.pipeshub.com | |
| PIPESHUB_CLIENT_SECRET | No | OAuth app client secret (used with client ID for OAuth authentication) |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
PipesHub MCP — instructions for the LLM
PipesHub is the user's workplace AI platform. It indexes their documents, knowledge base content, and connector sources (Drive, Box, Confluence, Slack, Jira, Gmail, ...). When in doubt, the answer is in PipesHub.
Full-document tasks: pipeshub_search → pipeshub_get_record_content
pipeshub_chat answers from a handful of retrieved passages — it never
reads a whole document. Whenever the task depends on a document's
COMPLETE content, fetch the document itself:
pipeshub_searchwith the document's name / topic.Take the top hit's
recordId.pipeshub_get_record_content, and answer from the returned content.
Tasks that need this path — anything where missing part of the document could make the answer wrong:
Summarize / TL;DR / key points / takeaways / action items of a doc.
Extract or list ALL of something (dates, owners, requirements, ...).
Check whether / where a doc mentions something.
Translate, rewrite, outline, review, or reformat a doc.
Compare named docs (fetch each
recordId).Any question explicitly scoped to ONE named document — chat retrieval cannot be restricted to a single record.
Default tool: pipeshub_chat
Use pipeshub_chat for any question that could plausibly be answered
by the user's PipesHub-indexed data and is not a full-document task:
A question that may span several documents, or where you don't yet know which record holds the answer (e.g. "what did Aashil say about onboarding?").
Anything about company / org policies, processes, decisions, or history (e.g. "what's our vacation policy?", "who owns the auth service?").
Anything explicitly mentioning PipesHub itself, or its sources (Drive / Box / Confluence / Slack / Gmail / Jira / etc.) when the user has those connected.
Open-ended "what do we know about X" questions where the answer likely lives in the org's documents.
Do NOT answer those from your own knowledge — pipeshub_chat grounds
the answer in the user's actual indexed content and returns citations
the user can verify.
When to use the other tools
pipeshub_search— locate a document by name or topic and resolve it to arecordId. To read or summarize one specific document, search, then pass the top hit'srecordIdtopipeshub_get_record_content.pipeshub_download_record— when the user wants the file as stored (download, attach, open). Never to read or summarize — that ispipeshub_get_record_contentmode:"content", which returns parsed text with a metadata header. Get therecordIdfrom a chat citation or frompipeshub_search.convertTois onlyapplication/pdf.pipeshub_get_record_content— when you need a record's full parsed content (returned as a singlecontentstring: metadata header plus the document's text) without downloading the original file. Prefer this over download when the question is about what the record says.pipeshub_directory— people, groups, teams, andwhoamilookups. Not for documents or files — that ispipeshub_search.pipeshub_sources— call once at the start of a session to discover which connectors / KB / models are available, then cache the result.pipeshub_agents— list the org's configured agents (specialized assistants with their own prompt, tools, and knowledge scope).
Knowledge Sources
A source (an app connector or a KB) holds record groups (a project, space, drive, folder), which hold records (one ticket, page, file). Records nest — an epic over its stories, a page over its child pages.
Every tool returns metadata or content.
pipeshub_search/pipeshub_chat— content, but only a ranked sample of blocks: never all blocks of a record, never every record that matches, never a complete list of what exists.mode:"lookup"andmode:"navigate"— metadata: recordIds, fields, and structure. No document text.mode:"content"— one whole record.
Combining them. Search finds a starting point by wording; navigate on a
recordId reveals what search structurally cannot — siblings, children,
linked records, and the group's total count. For a question that must be
exhaustive ("all", "how many", "every"), do NOT count search hits or trust a
chat answer: navigate the record group, which reports the real total. For a
record's siblings, navigate its parent — take the id from the Path line.
A named record — an issue key, a Confluence / Drive / Slack link, an
external ID → mode:"lookup" FIRST. It returns that record's metadata (for a
ticket: status, assignee, priority, dates) plus its recordId, which often
answers the question outright. Never guess a recordId.
For search, pass a connector id from pipeshub_sources as apps: ["<id>"]
and a collection id (kind: "knowledgeBase") as kb: ["<id>"]. Any source
id works as nodeId to navigate. Navigate and lookup end in a Next: line
naming the exact follow-up call — follow it.
Agents
Some orgs configure agents for specific jobs (e.g. a Slack messenger, a
Jira ticket creator, a Salesforce CRM updater). To run a turn against an
agent, call pipeshub_chat with its agentId (from pipeshub_agents)
— optionally with an agent chatMode (auto by default, or quick /
verification / deep). Keep passing the same agentId plus the
returned conversationId on follow-up turns.
When to route to an agent vs plain chat:
The user names a system/action that maps to an agent (e.g. "post to Slack", "create a Jira ticket", "update the Salesforce deal") → call
pipeshub_agents, pick the matching agent, and use it.Plain question about the org's knowledge or the public web → just use
pipeshub_chat(noagentId) withinternal_searchorweb_search.
If you are unsure which agent or route fits the request, call
pipeshub_agents FIRST and read the returned names/descriptions, then
either pick the best match or present the list to the user and let them
choose. Do NOT guess an agent blindly, and do NOT refuse — list the agents
and decide from real data. The list may be empty (no agents configured),
in which case fall back to plain pipeshub_chat.
Conversation lifecycle
pipeshub_chat is a single tool that handles both starting and
continuing conversations (plain or with an agentId). On the first turn
omit conversationId; on every subsequent turn pass back the
conversationId returned by the previous call (along with the same
agentId if you used one). Server-side context is preserved — do NOT
replay prior messages. Only start a fresh conversation when the user
explicitly asks to clear context.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| prompts | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| pipeshub_sourcesA | Discover available chat sources and AI models in one call. Returns up to three sections:
Call this once at the start of a session and cache the result — sources and models change infrequently. |
| pipeshub_chatA | Ask a question, get an answer grounded in the org's indexed data with citations. It reads a few retrieved passages — never a whole document, never a complete list. Three questions this tool gets WRONG. Check them first:
Everything else about the org's knowledge belongs here: policies, processes, decisions, history, "what do we know about X", and any question spanning several documents. Internal search (default, Web search ( Both are plain-chat modes. Agent chat — pass an
Conversation lifecycle — one tool, both start and continue:
Only re-omit The response contains the AI's |
| pipeshub_searchA | Vector / semantic search across the org's indexed documents. Use this when the user wants to LOCATE a document — by name, topic,
or a phrase to grep for — and to resolve it to a Typical uses:
Not for structural questions — what is under this epic, which pages are
in this space, what links to this ticket. Ranking by content cannot show
how records relate; use A ranked sample, never a complete list. Hits are the top-scoring blocks from the best-matching records — not all blocks of any record, and not every record that matches. Never count them to answer "how many" / "all" / "every"; navigate the record group instead, which reports its real total. By default it searches everything. To search only some sources, pass
connector ids in Each hit is one matching passage, best match first:
|
| pipeshub_download_recordA | Download the file as stored for one record — not PipesHub's parsed content, metadata header, or summary. Use this when the user wants the file itself (download, attach, open).
Get Do not use this to read, summarize, or answer "what does this doc
say?" regardless of format. That is
|
| pipeshub_get_record_contentA | Three operations on the org's records. Pick by what you hold:
Judge by the user's INTENT, not their keywords: "what's this doc about?",
"walk me through the report", "anything in here about Y?" are all
full-content tasks. Get the Returns one
Use it when the question depends on structure rather than wording: what is under this epic, which pages sit in this space, what is linked to this ticket, what is in this folder — and every "how many" / "all of" / "every" question. Search ranks by content; only this shows how records relate, and only this gives a count you can trust. Omit Pass Opening a record also prints that record's own metadata — for a ticket,
status, assignee, priority and dates — so a question about one record is
often answered by this call alone. It returns no document text; for that,
re-call with Returns
Handles Jira keys and URLs, Confluence, Drive, Slack permalinks, Linear, Notion, ServiceNow sys_id, SharePoint, Gmail/Outlook, and any connector whose records index a web URL. Resolution searches ALL connectors you can access, regardless of any source filter you used elsewhere. A miss is a 200 with empty Navigate and lookup return a rendered text view whose closing |
| pipeshub_directoryA | Look up people, groups, and teams in PipesHub. Five actions — pick
Omit |
| pipeshub_agentsA | List the PipesHub agents configured for this org, each with its capabilities. Agents are specialized assistants (custom system prompt, tools, knowledge
scope). To converse with one, take its Each agent is returned as:
Route on The list may be empty (no agents configured). For plain Q&A when no
specific agent is needed, use |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| pipeshub-assistant | Load PipesHub tool-routing guidance so the assistant picks the correct tool (pipeshub_chat / search / agents / directory) and lists agents with pipeshub_agents when unsure which route fits the query. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 7 tools
Each tool targets a distinct operation: search locates records, chat answers grounded questions, get_record_content reads full text/navigates/looks up, download_record fetches original bytes, and agents/directory/sources handle their own domains. The descriptions explicitly cross-reference and warn against misuse, so an agent can reliably select the right tool.
All tools share the pipeshub_ prefix, but the pattern after it is inconsistent: some are verb_noun (download_record, get_record_content), some are bare verbs (search, chat), and some are bare nouns (agents, directory, sources). This makes names individually readable but not predictably derivable.
Seven tools is a well-scoped size for a knowledge retrieval and chat server. Each tool covers a major capability without unnecessary fragmentation, and no tool feels redundant.
The surface covers the full read-side workflow: discover sources, search, chat with citations, resolve external references, read full content, navigate hierarchy, download files, and look up agents and people. There are no obvious gaps or dead ends for the server's apparent purpose.