konsol-mcp
MCP server for configuring the konsol Frappe app via the CLI. Enables AI clients to manage dimensions, measures, fact tables, connectors, and schemas using natural language.
README
konsol-cli
Command-line client for configuring the konsol Frappe app.
All configuration goes through konsol — never dbt or ClickHouse directly.
Install
cd konsol_cli
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
Optional MCP server:
pip install -e ".[mcp]"
Two ways to connect
1. Local Docker (default)
Talks to konsol by running bench execute inside your Docker stack.
konsol dimension list --compose-file ../repo/docker-compose.yml
2. Remote HTTP API
Talks to any running konsol site over the Frappe API. Use this from your laptop against a remote server, in CI, or with MCP.
First, create API keys in Frappe:
- Log into the site as an EPM Admin user
- Go to User → your user → Settings tab → API Access
- Click Generate Keys and save the key + secret
Then run commands:
konsol --backend api \
--url http://localhost:8069 \
--site konsolidat.local \
--api-key YOUR_KEY \
--api-secret YOUR_SECRET \
dimension list
When connecting to localhost:8069, keep --site set — the CLI sends it as the Host header so Frappe routes to the right site.
Optional config file — copy config.example.toml to ~/.config/konsol/config.toml (connection settings only).
Credentials (pick one, never commit):
cp secrets.example.env ~/.config/konsol/secrets.env && chmod 600 ~/.config/konsol/secrets.env
# or: secrets.example.toml → ~/.config/konsol/secrets.toml
The CLI loads secrets.env first, then secrets.toml. Shell env vars take precedence.
Grok MCP — configs use ${KONSOL_API_KEY} placeholders. Before starting Grok:
source scripts/load-konsol-secrets.sh # exports vars, does not print them
grok mcp doctor konsol
See .grok/config.toml (project-scoped, safe to commit) — not plaintext secrets.
Commands
# Dimensions
konsol dimension list
konsol dimension show dim_cost_center
konsol dimension create dim_project --source-column Project --label "Project"
konsol dimension publish dim_project
konsol dimension unpublish dim_project
# Measures
konsol measure list
konsol measure create period_headcount --expression "sum(headcount)" --label "Headcount"
konsol measure unpublish period_headcount
# Fact tables
konsol fact list
konsol fact show headcount
konsol fact create my_fact --label "My Fact" --source-type Statistical \
--clickhouse-table epm_staging.fact_x --scenario-key statistical
konsol fact publish my_fact
konsol fact unpublish my_fact
# Connectors & ERP sources
konsol connector list
konsol connector show CONN-00001
konsol connector create --name "ERPNext Demo" --erp-type erpnext --entity-id ENT01
konsol connector delete CONN-00001
konsol source list
# Config (GitOps YAML)
konsol config export -o model.yaml
konsol config diff model.yaml
konsol config apply model.yaml [--publish] [--prune] [--dry-run]
# Schema
konsol schema apply
konsol schema status
MCP (AI clients)
Install the optional extra, then add mcp.example.json to your MCP client config (Cursor, Claude Desktop, etc.):
pip install -e ".[mcp]"
Set KONSOL_URL, KONSOL_SITE, KONSOL_API_KEY, and KONSOL_API_SECRET in the server env block, then run konsol-mcp (stdio transport).
Connector Airbyte tools (v0.10.0+): test_connector_writeback, provision_connector_airbyte — require the matching konsol app endpoints on your site.
Debug with the MCP Inspector:
npx @modelcontextprotocol/inspector konsol-mcp
Environment variables
| Variable | Purpose |
|---|---|
KONSOL_BACKEND |
bench (default) or api |
KONSOL_SITE |
Frappe site name |
KONSOL_URL |
Site URL for api backend |
KONSOL_API_KEY |
Frappe API key |
KONSOL_API_SECRET |
Frappe API secret |
KONSOL_COMPOSE_FILE |
Path to docker-compose.yml for bench backend |
Docker deploy note
bench get-app installs konsol from git HEAD, not uncommitted files. Commit konsol changes before docker compose build frappe_backend.
Architecture
konsol-cli / konsol-mcp
→ bench execute or HTTP /api/method/konsol.cli_api.*
→ konsol.config_service → Frappe DocTypes
→ publish / apply_schema → dbt vars, ClickHouse DDL, Pipeline Build Request
Tests
cd konsol_cli && .venv/bin/python -m pytest tests/ -q
cd repo/docker/frappe/konsol && PYTHONPATH=konsol .venv/bin/python -m pytest konsol/tests/test_config_service.py -q
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。