kdeps

About kdeps

kdeps is an open-source framework (Apache 2.0) for building, orchestrating, and deploying AI agents entirely through declarative YAML — no Python scripts, no Flask wiring, no manual dependency chains.

Three run modes: - Workflow — Deterministic DAG pipelines. Compose resources (LLM calls, HTTP, SQL, Python, scraping, search, email, browser, git, code intelligence) via requires: dependencies. Runs as an API server, web server, file watcher, or Telegram bot. - Agent — Autonomous LLM loop with tool use. Each workflow becomes a callable tool; the model decides which tools to invoke and in what order. REPL with slash commands, session persistence (PostgreSQL, MongoDB, SQLite), and resume support. - Agency — Multi-agent orchestration. Agents call other agents via the agent: resource type, passing output as params. Like calling a function, but each function is an entire AI pipeline.

Capabilities: - 17+ resource types: chat, httpClient, python, exec, sql, email, scraper, browser, embedding, searchLocal, searchWeb, agent, component, file, git, codeIntelligence - Multi-backend LLM: local llamafile (zero install), Ollama, OpenAI, Anthropic, Groq, xAI, Perplexity, OpenRouter, and any OpenAI-compatible endpoint - 16 vector store backends (Qdrant, Weaviate, pgvector, MongoDB, MySQL/MariaDB/Dolt, AlloyDB, CloudSQL, Milvus, Chroma, LanceDB, and more) - Built-in tools: calculator (Starlark math), bash_exec, Wolfram Alpha, Cohere/Voyage/Jina rerank, Zapier NLA - Export to Docker image, Kubernetes manifests, bootable ISO, or self-contained binary - Component registry at kdeps.io — install pre-built packages with one command - Structured JSON logging, TLS, rate limiting, CORS, authentication - Jinja2 expressions for data flow between resources (get(), output(), set())

Install in one command: curl -LsSf https://raw.githubusercontent.com/kdeps/kdeps/main/install.sh | sh

From idea to running API in three steps: write YAML → run kdeps → call the endpoint.

Keywords

• Agentic Ai
• Ai
• Ai Agents
• Developer Tools
• Llm Orchestration
• Mlops
• Multi-Agent Systems
• Open Source
• Rag
• Rag Pipelines
• Vector Stores
• Workflow Automation
• Yaml Configuration

Address

• Headquarters

  Ouddiemerlaan 3091111HA DiemenNoord-Holland, Netherlands

Team by function

Based on 1 public job profiles as of 3 September 2026.

• Engineering / Software Development1