MCP Connection Guide
Connect any AI tool to your Pidge workspace.
Quick Connect
Set the MCP server URL
Point your client at https://getpidge.com/mcp.
Complete OAuth in the browser
Clients should start authorization at https://getpidge.com/oauth/authorize.
Approve only the read scopes you need
Pack access is read-only. MCP clients call tools for packs, memory, files, search, and tasks based on granted scopes.
Use SSE only for legacy clients
Modern clients should use https://getpidge.com/mcp; legacy SSE clients can fall back to https://getpidge.com/sse.
Use the top-level references for discovery
Machine-readable docs are published at https://getpidge.com/agents.md and https://getpidge.com/llms.txt.
Endpoints
MCP Server
https://getpidge.com/mcp
OAuth Authorization
https://getpidge.com/oauth/authorize
Authorization Metadata
https://getpidge.com/.well-known/oauth-authorization-server
Protected Resource Metadata
https://getpidge.com/.well-known/oauth-protected-resource
MCP Resource Metadata
https://getpidge.com/.well-known/oauth-protected-resource/mcp
SSE Resource Metadata
https://getpidge.com/.well-known/oauth-protected-resource/sse
Legacy SSE
https://getpidge.com/sse
Top-Level agents.md
https://getpidge.com/agents.md
Top-Level llms.txt
https://getpidge.com/llms.txt
Well-Known llms.txt
https://getpidge.com/.well-known/llms.txt
Available Scopes
read:memory
Read daily logs, pinned facts, search memory
write:memory
Create/update pinned facts, add daily entries
read:packs
List and read installed packs and marketplace packs
read:files
List and read virtual filesystem
write:files
Write/delete files in virtual filesystem
read:search
Full-text and semantic search
read:tasks
List and check task status
write:tasks
Start new tasks
Available Tools
memory_search
Search business memory using natural language and return follow-up IDs.
Required scope: read:memory
Parameters: { query: string, limit?: number }
memory_read
Read a specific memory entry by ID from memory_search results.
Required scope: read:memory
Parameters: { id: string }
memory_daily
Read daily log entries by date or range.
Required scope: read:memory
Parameters: { date?: string, days?: number }
memory_pinned
List pinned facts and long-lived knowledge.
Required scope: read:memory
Parameters: { category?: string }
memory_remember
Pin a new fact to persistent memory.
Required scope: write:memory
Parameters: { content: string, category?: string }
search
Run hybrid keyword and semantic search across indexed content.
Required scope: read:search
Parameters: { query: string, limit?: number, scope?: "all"|"memory"|"files"|"packs" }
files_list
List files in the virtual filesystem.
Required scope: read:files
Parameters: { path?: string, recursive?: boolean }
files_read
Read file contents from the virtual filesystem.
Required scope: read:files
Parameters: { path: string }
files_write
Write or overwrite a file in the virtual filesystem.
Required scope: write:files
Parameters: { path: string, content: string }
packs_list
List installed packs or browse the public Pidge marketplace.
Required scope: read:packs
Parameters: { source?: "installed"|"marketplace", query?: string, category?: string, tag?: string, limit?: number, offset?: number }
packs_read
Read a specific installed pack or marketplace pack in detail.
Required scope: read:packs
Parameters: { name: string, source?: "installed"|"marketplace" }
tasks_list
List tasks with optional status filtering.
Required scope: read:tasks
Parameters: { status?: "pending"|"running"|"completed"|"failed", limit?: number }
task_start
Start a new background task from a natural-language instruction.
Required scope: write:tasks
Parameters: { instruction: string, priority?: "low"|"normal"|"high" }
See Agent Integration for the machine-readable reference.
Launch Contract
Phase 39 launch contract is tools-only. MCP resources and prompts are not part of the supported public surface.
Transport Notes
Use https://getpidge.com/mcp for streamable HTTP clients. The legacy SSE transport at https://getpidge.com/sse stays available for older clients: open a long-lived GET stream first, then POST JSON-RPC messages to the session endpoint from the initial endpoint event.