Using Continue with mittwald
Continue is an open-source AI code assistant that works with both JetBrains IDEs and Visual Studio Code.
Installation
Install the Continue plugin for your IDE:
- JetBrains IDEs: JetBrains Marketplace
- Visual Studio Code: VS Code Marketplace
Configuration
Continue is configured via a config.yaml file, typically located in ~/.continue/ in your home directory.
Basic Configuration
Add mittwald as a model provider:
name: mittwald
version: 1.0.0
schema: v1
models:
- name: DeepSeek-V4.1-Flash-1017
provider: openai
model: DeepSeek-V4.1-Flash-1017
apiBase: https://llm.aihosting.mittwald.de/v1
apiKey: sk-your-api-key-here
roles:
- chat
- edit
- apply
capabilities:
- tool_use
- image_input
defaultCompletionOptions:
contextLength: 1048576
maxTokens: 32768
temperature: 1.0
topP: 1.0
- name: Kolibri-1
provider: openai
model: Kolibri-1
apiBase: https://llm.aihosting.mittwald.de/v1
apiKey: sk-your-api-key-here
roles:
- chat
defaultCompletionOptions:
contextLength: 1048576
maxTokens: 32768
temperature: 1.0
topP: 0.97
topK: 128
context:
- provider: code
- provider: docs
- provider: diff
- provider: terminal
- provider: problems
- provider: folder
- provider: codebase
Replace sk-your-api-key-here with your mittwald AI Hosting API key (how to get one).
The configuration adds two models:
- DeepSeek-V4.1-Flash-1017 for chat, edits, and Agent mode. It reads images and reasons
before it answers, so
maxTokensleaves room for the reasoning. If an answer stops before the model writes anything, raisemaxTokens. - Kolibri-1 for chat in German. It reads text only and calls tools only with
tool_choice: "auto", so the configuration leaves outtool_use. Use DeepSeek-V4.1-Flash-1017 in Agent mode.
Both models are in the Testphase. For production use, choose Qwen3.8-27B-NVFP4
(Stable) with contextLength: 256000 instead.
MCP Server Integration
Continue supports Model Context Protocol servers for enhanced capabilities. Add MCP servers to your config.yaml:
mcpServers:
# Context7: Up-to-date library documentation
context7:
name: Context7
command: npx
args: ["-y", "@upstash/context7-mcp"]
env: { CONTEXT7_API_KEY: "your-context7-api-key" }
# Brave Search: Web search with privacy focus
brave-search:
name: Brave
command: npx
args: ["-y", "@brave/brave-search-mcp-server"]
env: { BRAVE_API_KEY: "your-brave-api-key" }
# Directus: Headless CMS data management
directus:
name: Directus
command: npx
args: ["-y", "@directus/content-mcp", "--stdio"]
env:
DIRECTUS_URL: "https://your-directus-instance.com"
DIRECTUS_TOKEN: "your-directus-token"
# OpenSearch: Search and analytics engine
opensearch:
name: OpenSearch
command: uvx
args: ["opensearch-mcp-server-py"]
env:
OPENSEARCH_URL: "https://localhost:9200"
OPENSEARCH_USERNAME: "your-opensearch-username"
OPENSEARCH_PASSWORD: "your-opensearch-password"
# TYPO3: CMS content and backend management (requires mittwald CLI)
typo3:
name: TYPO3
command: mw
args:
[
"app",
"exec",
"--quiet",
"--installation-id=a-XXXXX",
"vendor/bin/typo3 mcp:server",
]
# mittwald MCP: Infrastructure management
mittwald:
name: mittwald
transport:
type: sse
url: https://mcp.mittwald.de/mcp
Using MCP Tools
Ask the assistant explicitly to use a tool:
- "Use Brave Search to find recent information about React 19..."
- "Query OpenSearch for recent error logs..."
- "Use mittwald MCP to list my projects"
Always scope access narrowly and review any data access or modifications.
Best Practices
Safety with MCP Servers
- Scope access narrowly (specific databases, domains, paths)
- Log all tool calls for audit purposes
- Use low temperature (≤ 0.2) for tool planning to reduce hallucinations. Reasoning models such as DeepSeek-V4.1-Flash-1017 are the exception: keep the sampling parameters from their model pages
- Review diffs and queries before executing
- Start with read-only access when possible
Next Steps
- Explore other IDE integrations - Try different tools
- Learn about MCP - Deep dive into infrastructure management
- Install agent skills - Add workflow guidance