Complete reference for all 9 MCP tools exposed by the Loop MCP server. Covers dispatch, issue management, signals, comments, and dashboard.

Tools

The Loop MCP server exposes 9 tools that give AI agents full access to the Loop feedback loop. Tools are grouped into four categories: dispatch, issue management, signals, and observability.

Get the highest-priority unblocked issue with dispatch instructions. Atomically claims the issue so no other agent can pick it up simultaneously. Returns the issue details along with a hydrated prompt containing full execution instructions.

When the dispatch queue is empty, returns a message indicating no tasks are available.

Parameters

Name	Type	Required	Description
`projectId`	`string`	No	Filter to a specific project

Example response

{
  "issue": {
    "id": "clx1abc2def3ghi4jkl5mnop",
    "number": 42,
    "title": "Fix OAuth redirect showing blank page",
    "type": "task",
    "priority": 1,
    "status": "in_progress"
  },
  "prompt": "You are working on issue #42...",
  "meta": {
    "templateSlug": "signal-triage",
    "templateId": "tpl_abc...",
    "versionId": "ver_xyz...",
    "versionNumber": 1,
    "reviewUrl": "POST /api/prompt-reviews"
  }
}

loop_complete_task

Mark an issue as done, record what was accomplished, and find newly unblocked issues. This tool performs three actions in sequence:

Sets the issue status to done
Posts an outcome comment attributed to the agent
Queries for issues that were blocked by the completed one

Parameters

Name	Type	Required	Description
`issueId`	`string`	Yes	ID of the issue to complete
`outcome`	`string`	Yes	Summary of what was accomplished

Example response

{
  "issue": { "id": "clx1abc...", "status": "done" },
  "comment": { "id": "clx2def...", "body": "Fixed the blank page..." },
  "unblockedIssues": [
    { "id": "clx3ghi...", "title": "Add latency tracking" }
  ]
}

Issue Management

loop_create_issue

Create a new issue in Loop. Issues are the atomic unit of work -- signals, hypotheses, plans, tasks, and monitors are all issue types.

Parameters

Name	Type	Required	Description
`title`	`string`	Yes	Issue title (1-500 characters)
`type`	`string`	No	One of `signal`, `hypothesis`, `plan`, `task`, `monitor`. Defaults to `task`
`priority`	`number`	No	`0` = none, `1` = urgent, `2` = high, `3` = medium, `4` = low. Defaults to `0`
`projectId`	`string`	No	Project to assign the issue to
`description`	`string`	No	Issue description
`parentId`	`string`	No	Parent issue ID for hierarchy

Example response

{
  "id": "clx1abc2def3ghi4jkl5mnop",
  "number": 43,
  "title": "Add loading spinner to OAuth redirect",
  "type": "task",
  "status": "triage",
  "priority": 2
}

loop_update_issue

Update an existing issue in Loop. You can change the status, priority, type, title, or description. Only the fields you provide will be updated.

Parameters

Name	Type	Required	Description
`issueId`	`string`	Yes	ID of the issue to update
`status`	`string`	No	One of `triage`, `backlog`, `todo`, `in_progress`, `done`, `canceled`
`priority`	`number`	No	`0` = none, `1` = urgent, `2` = high, `3` = medium, `4` = low
`type`	`string`	No	One of `signal`, `hypothesis`, `plan`, `task`, `monitor`
`title`	`string`	No	New title (1-500 characters)
`description`	`string`	No	New description

Example response

{
  "id": "clx1abc2def3ghi4jkl5mnop",
  "number": 42,
  "title": "Fix OAuth redirect showing blank page",
  "type": "task",
  "status": "in_progress",
  "priority": 1
}

loop_list_issues

List issues from Loop with optional filters for status, type, and project. Returns paginated results with a hasMore flag for cursor-based iteration.

Parameters

Name	Type	Required	Description
`status`	`string`	No	Filter by status: `triage`, `backlog`, `todo`, `in_progress`, `done`, `canceled`
`type`	`string`	No	Filter by type: `signal`, `hypothesis`, `plan`, `task`, `monitor`
`projectId`	`string`	No	Filter by project ID
`limit`	`number`	No	Maximum issues to return (1-100, default 20)
`offset`	`number`	No	Number of issues to skip for pagination

Example response

{
  "issues": [
    {
      "id": "clx1abc...",
      "number": 42,
      "title": "Fix OAuth redirect",
      "type": "task",
      "status": "todo",
      "priority": 1
    }
  ],
  "total": 128,
  "hasMore": true
}

loop_get_issue

Get full details of a single issue by ID, including labels, comments, and relations. Use this to inspect an issue before working on it or to check the current state after updates.

Parameters

Name	Type	Required	Description
`issueId`	`string`	Yes	ID of the issue to retrieve

Example response

{
  "id": "clx1abc2def3ghi4jkl5mnop",
  "number": 42,
  "title": "Fix OAuth redirect showing blank page",
  "type": "task",
  "status": "in_progress",
  "priority": 1,
  "description": "Login requests are timing out...",
  "labels": [{ "id": "clx4...", "name": "bug" }],
  "comments": [{ "id": "clx5...", "body": "Investigating..." }],
  "relations": [{ "id": "clx6...", "type": "blocked_by", "targetId": "clx7..." }]
}

Signals

loop_ingest_signal

Ingest a signal (error, metric change, user feedback) into Loop. Creates a signal record and a linked triage issue automatically. This is how agents report observations back into the feedback loop.

Parameters

Name	Type	Required	Description
`source`	`string`	Yes	Signal source (e.g., `"agent"`, `"posthog"`, `"sentry"`)
`type`	`string`	Yes	Signal type (e.g., `"error"`, `"metric_change"`, `"user_feedback"`)
`severity`	`string`	No	One of `critical`, `high`, `medium`, `low`. Defaults to `medium`
`payload`	`object`	Yes	Signal data as key-value pairs
`projectId`	`string`	No	Project to associate the signal with

Example response

{
  "signal": {
    "id": "clx8sig...",
    "source": "agent"
  },
  "issue": {
    "id": "clx9iss...",
    "number": 44,
    "title": "Signal: error from agent",
    "status": "triage"
  }
}

Comments

loop_create_comment

Post a comment on an issue as the agent. Comments are automatically attributed with authorType: "agent". Use this to report progress, document findings, or ask questions on an issue.

Parameters

Name	Type	Required	Description
`issueId`	`string`	Yes	ID of the issue to comment on
`body`	`string`	Yes	Comment text (minimum 1 character)

Example response

{
  "id": "clx2def3ghi4jkl5mnop6qrs",
  "body": "Completed investigation. The root cause is...",
  "authorName": "agent",
  "createdAt": "2026-02-23T12:00:00.000Z"
}

{
  "issues": {
    "total": 128,
    "byStatus": {
      "triage": 12,
      "backlog": 34,
      "todo": 28,
      "in_progress": 8,
      "done": 42,
      "canceled": 4
    }
  },
  "goals": { "active": 3, "completed": 7 },
  "dispatch": { "queueDepth": 28 }
}

Tool behavior notes

Error handling

All tools return structured error messages when something goes wrong. Errors include an HTTP status code and a human-readable message. Common errors:

Status	Meaning
`401`	Invalid or missing API key
`404`	Issue or resource not found
`422`	Validation error (bad input)
`500`	Server error

Read-only vs mutating tools

Tool	Read-only	Idempotent
`loop_get_next_task`	No	No
`loop_complete_task`	No	Yes
`loop_create_issue`	No	No
`loop_update_issue`	No	No
`loop_list_issues`	Yes	Yes
`loop_get_issue`	Yes	Yes
`loop_ingest_signal`	No	No
`loop_create_comment`	No	No
`loop_get_dashboard`	Yes	Yes

Typical agent workflow

A standard agent loop uses these tools in sequence:

loop_get_next_task  ->  (do work)  ->  loop_complete_task
       |                                      |
       |                                      v
       |                            loop_create_comment
       |                            loop_ingest_signal
       |                            loop_create_issue
       v
loop_get_dashboard  (check system health)

Call loop_get_next_task to claim the highest-priority issue
Read the hydrated prompt and execute the work
Use loop_create_comment to report progress
If new issues are discovered, use loop_create_issue to add them
If signals are observed (errors, metrics), use loop_ingest_signal
Call loop_complete_task with an outcome summary when done
Optionally call loop_get_dashboard to check overall system health

Tools

Tools

Dispatch

loop_get_next_task

Parameters

Example response

loop_complete_task

Parameters

Example response

Issue Management

loop_create_issue

Parameters

Example response

loop_update_issue

Parameters

Example response

loop_list_issues

Parameters

Example response

loop_get_issue

Parameters

Example response

Signals

loop_ingest_signal

Parameters

Example response

Comments

loop_create_comment

Parameters

Example response

Observability

loop_get_dashboard

Parameters

Example response

Tool behavior notes

Error handling

Read-only vs mutating tools

Typical agent workflow

On this page