LABSCO SUMMARY

The 'why did last night fail' loop closes in one place: failed_dags, then the task, then its log.

The read surface is wide enough that most incident questions never reach the web UI, and the write surface is deliberately small — trigger, pause, unpause, and connection create, update and delete. The thing to check first is version: list_users, list_roles, list_permissions, list_datasets and the dataset event tools are marked v1 API only, so on a v2 cluster those are the calls that will not behave, while DAG, task instance and config coverage is unaffected.

What it is

A server over the Airflow REST API, 54 tools spanning DAGs and their runs, task instances and their logs, configuration, pools, variables, connections, users and roles, plugins, providers and datasets.

What you get

The triage path in three steps: running_dags and failed_dags say what is broken now, dag_task_duration narrows it to a task, get_task_instance_logs returns that attempt's log by try_number
DAG control alongside the reads — trigger_dag, pause_dag and unpause_dag — with dag_code, dag_graph and list_tasks to read the DAG itself
Import errors treated as their own view: list_import_errors, get_import_error and all_dag_import_summary, so a DAG that never registered is visible next to the ones that ran
Listing shaped for a large cluster: list_dags and get_dags_detailed_batch take name_contains and id_contains filters with offset and limit, or fetch_all in one go
Cluster state and settings: get_health, get_version, get_config with list_config_sections and search_config_options, plus pools, variables and event logs
Connections that can be written, not only read — create_connection, update_connection and delete_connection sit beside get_connection
Values passed between tasks and data lineage: list_xcom_entries and get_xcom_entry, list_datasets and list_dataset_events

Requirements

An Airflow cluster reachable over its REST API, with an account whose role permits the endpoints in play — the connection tools write to it. The user, role, permission and dataset tools are documented as v1 API only, and the dataset ones note that v2 uses Assets instead. MIT licensed.

Setup effort

One command plus credentials — uvx --python 3.12 mcp-airflow-api, then supply credentials

See installation options

USE CASES

01

Working out why last night's run failed

Start from the failed runs rather than the DAG list, compare task durations inside the run that broke, open the failing task instance and read the log for the attempt that matters.

failed_dagsdag_task_durationget_task_instance_detailsget_task_instance_logs

02

Finding the DAG that never showed up

A DAG that fails to import never appears in the run history at all. The import error tools show it directly, with the summary telling you whether it is one broken file or a shared import that took several DAGs down with it.

list_import_errorsget_import_errorall_dag_import_summary

03

Holding the schedule during a maintenance window

Check what is in flight before touching anything, pause the DAGs that must not start, then unpause them and trigger a catch-up run once the window closes.

running_dagspause_dagunpause_dagtrigger_dag

04

Tracking down a pipeline that keeps getting slower

Compare run durations over recent history, then break the worst run down by task, and pull the task instances across a date and state range to see whether the delay is one operator or queueing.

dag_run_durationdag_task_durationlist_task_instances_all

05

Repointing a DAG at a new database

Find the connection the DAG uses, read what it currently holds, and update it in place — the connection tools write, so the change does not need a trip to the UI.

list_connectionsget_connectionupdate_connection

54 tools

Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.

`all_dag_event_summary`

[Tool Role]: Provides summary of event logs across all DAGs.

BEFORE YOU INSTALL

What to know first

Tools you get

54 tools when we connected. The documented totals of 43 for API v1 and 45 for v2 are out of date. AIRFLOW_API_VERSION still changes the set: the README marks the asset tools, list_assets and list_asset_events, as API v2 only.

Requires

uvx on PATH
AIRFLOW_API_PASSWORD in your environment
AIRFLOW_API_USERNAME in your environment

CONTEXT COST

Moderate Standing cost

Rated by the size of the 54 tool definitions that ride along on every turn, not by how many there are

INSTALLATION

Connect your client

Server setup

TERMINAL

uvx --python 3.12 mcp-airflow-api

MCP Clients

Claude Desktop — local stdio

CLAUDE DESKTOP — LOCAL STDIO · JSON

{
  "mcpServers": {
    "mcp-airflow-api": {
      "command": "uvx",
      "args": [
        "--python",
        "3.12",
        "mcp-airflow-api"
      ],
      "env": {
        "AIRFLOW_API_VERSION": "v2",
        "AIRFLOW_API_BASE_URL": "http://localhost:8080/api",
        "AIRFLOW_API_USERNAME": "airflow",
        "AIRFLOW_API_PASSWORD": "airflow"
      }
    }
  }
}

Any MCP client — HTTP

ANY MCP CLIENT — HTTP · JSON

{
  "mcpServers": {
    "mcp-airflow-api": {
      "type": "streamable-http",
      "url": "http://localhost:8000/mcp"
    }
  }
}

Clients spell the "type" value differently: Claude Code and VS Code use "http", Cline uses "streamableHttp", and Roo Code uses "streamable-http". Claude Desktop does not accept a URL entry in its config file.

Any MCP client — HTTP, bearer token

ANY MCP CLIENT — HTTP, BEARER TOKEN · JSON

{
  "mcpServers": {
    "mcp-airflow-api": {
      "type": "streamable-http",
      "url": "http://localhost:8000/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_SECURE_SECRET_KEY>"
      }
    }
  }
}

Clients spell the "type" value differently: Claude Code and VS Code use "http", Cline uses "streamableHttp", and Roo Code uses "streamable-http". Claude Desktop does not accept a URL entry in its config file.

Open WebUI — Docker Compose

OPEN WEBUI — DOCKER COMPOSE · TERMINAL

git clone https://github.com/call518/MCP-Airflow-API.git
cd MCP-Airflow-API
cp .env.example .env

Edit .env with your Airflow API settings

OPEN WEBUI — DOCKER COMPOSE · TERMINAL

docker-compose up -d

Open OpenWebUI at http://localhost:3002/

Airflow-API MCP

The 'why did last night fail' loop closes in one place: failed_dags, then the task, then its log.

What it is

What you get

Requirements

Setup effort

Working out why last night's run failed

Finding the DAG that never showed up

Holding the schedule during a maintenance window

Tracking down a pipeline that keeps getting slower

Repointing a DAG at a new database

54 tools

all_dag_event_summary

all_dag_import_summary

create_connection

dag_calendar

dag_code

dag_graph

dag_run_duration

dag_task_duration

delete_connection

failed_dags

get_config

get_config_section

get_connection

get_dag

get_dags_detailed_batch

get_dataset

get_dataset_events

get_event_log

get_health

get_import_error

get_pool

get_prompt_template

get_provider

get_task_instance_details

get_task_instance_extra_links

get_task_instance_logs

get_user

get_variable

get_version

get_xcom_entry

list_config_sections

list_connections

list_dags

list_dataset_events

list_datasets

list_event_logs

list_import_errors

list_permissions

list_plugins

list_pools

list_providers

list_roles

list_task_instances_all

list_task_instances_batch

list_tasks

list_users

list_variables

list_xcom_entries

pause_dag

running_dags

search_config_options

trigger_dag

unpause_dag

update_connection

What to know first

Tools you get

Requires

Moderate Standing cost

Connect your client

Server setup

MCP Clients

Claude Desktop — local stdio

Any MCP client — HTTP

Any MCP client — HTTP, bearer token

Open WebUI — Docker Compose

Same category

Kafka MCP

Lenses

`all_dag_event_summary`

`all_dag_import_summary`

`create_connection`

`dag_calendar`

`dag_code`

`dag_graph`

`dag_run_duration`

`dag_task_duration`

`delete_connection`

`failed_dags`

`get_config`

`get_config_section`

`get_connection`

`get_dag`

`get_dags_detailed_batch`

`get_dataset`

`get_dataset_events`

`get_event_log`

`get_health`

`get_import_error`

`get_pool`

`get_prompt_template`

`get_provider`

`get_task_instance_details`

`get_task_instance_extra_links`

`get_task_instance_logs`

`get_user`

`get_variable`

`get_version`

`get_xcom_entry`

`list_config_sections`

`list_connections`

`list_dags`

`list_dataset_events`

`list_datasets`

`list_event_logs`

`list_import_errors`

`list_permissions`

`list_plugins`

`list_pools`

`list_providers`

`list_roles`

`list_task_instances_all`

`list_task_instances_batch`

`list_tasks`

`list_users`

`list_variables`

`list_xcom_entries`

`pause_dag`

`running_dags`

`search_config_options`

`trigger_dag`

`unpause_dag`

`update_connection`