MCP Overview

Connect AI agents to Allye through the hosted remote MCP server at https://mcp.allye.app/mcp.

Allye MCP is a hosted server that implements the Model Context Protocol. It gives an AI agent a structured way to work with your Allye workspace: read projects, epics, specs and tasks, recall decisions from memories, use the knowledge base, and apply shared skills.

The interface is organized around intent. The agent picks the capability it needs, then selects an action inside it. It does not need to know how Allye stores the data behind the interface.

What MCP enables

With Allye connected, an agent can:

load the user, team and workspace context before starting work;
read specs and tasks, start a task, and submit it for review;
save durable decisions as memories and search them semantically;
create and navigate knowledge base pages;
list, read, create and edit skills, and review or restore their revisions;
search the workspace API catalog for product context.

How the connection works

Agent (Claude Code, Codex, OpenCode, Pi, OMP)
          │
          │ https://mcp.allye.app/mcp
          │ OAuth 2.1
          ▼
   Hosted Allye MCP server
          │
          ├── Projects, Epics, Specs, Tasks
          ├── Memories (intelligence)
          ├── Knowledge base (kb)
          ├── Skills
          ├── Teams, TODOs, user config
          └── API catalog

Allye MCP is offered only as a hosted remote service at https://mcp.allye.app/mcp. Nothing is installed or started on your machine. The client opens a browser window, you approve access with OAuth 2.1, and the client manages the session and reconnection. No personal access token is involved.

Supported harnesses

Harness	How to connect
Claude Code	allye-plugin or the remote MCP URL
Codex	Install guide
OpenCode	Install guide
Pi	Install guide
OMP	Install guide

See Installation & Setup for the steps.

Capabilities by intent

Intent	MCP tool	Typical use
Start or refresh context	`initialize`	Load user, workspace, team and document context.
Organize delivery	`projects`, `epics`	List and edit projects and epics; manage the project's apps.
Specify work	`specs`	Read spec context and anchors, submit, approve, track versions.
Execute work	`tasks`	List your next task, start, submit, complete or request changes.
Remember decisions	`intelligence`	Save, search, traverse and relocate semantic memories.
Write and organize knowledge	`kb`	Search, read and write knowledge base pages in company, project and personal spaces; sections, backlinks, history and restore.
Reuse agent guidance	`skills`	List, read, create and update skills; list and restore revisions.
Manage personal follow-ups	`productivity`	Create, search and update personal TODOs.
Work with teams	`team`	List teams, set the default team, inspect members.
Personal configuration	`user_config`	Create and maintain user configuration documents.
Find product API knowledge	`api_catalog`	Search and inspect catalogued API specifications.
Check availability	`allye_health_check`	Confirm the service is responding.

Action names for each tool are listed in the Actions Reference.

Protocol capabilities

Allye MCP exposes MCP tools, prompts and resources. Prompts cover the API catalog, your open work, memories and TODOs. The allye://my-teams resource returns the teams available to the authenticated user.

Teams and context

A connection authenticates you and establishes the workspace context. You can belong to more than one team. Project, epic, spec and task calls follow the team of the project they address. Use team with team_list to see your teams and team_set_default to change your default team, which is used to create projects and for memory, TODO and skill calls that do not name a team.

A typical interaction

The client performs the MCP initialize exchange automatically.
The agent reads your teams and default team from the context, and passes team_id when the default is not the right one.
The agent calls a tool with tools/call and an action-specific argument object.
The agent uses the result to answer you or continue the workflow.

For example, a semantic memory lookup is a plain tool call:

{
  "name": "intelligence",
  "arguments": {
    "action": "memory_search",
    "query": "decisions about authentication",
    "limit": 5
  }
}