██╗ ██╗███████╗ ██████╗████████╗ ██████╗ ██████╗ ██║ ██║██╔════╝██╔════╝╚══██╔══╝██╔═══██╗██╔══██╗ ███████║█████╗ ██║ ██║ ██║ ██║██████╔╝ ██╔══██║██╔══╝ ██║ ██║ ██║ ██║██╔══██╗ ██║ ██║███████╗╚██████╗ ██║ ╚██████╔╝██║ ██║ ╚═╝ ╚═╝╚══════╝ ╚═════╝ ╚═╝ ╚═════╝ ╚═╝ ╚═╝ 

Config-first A2A-Native Agent Platform

Deploy observable, secure, and scalable AI agents with unified configuration (YAML), plus a programmatic API.

Documentation | Quick Start | Configuration

The native desktop GUI for Hector. Manage your agents and workspaces with a rich visual interface.

• Project: verikod/hector-studio
• Download: Latest Releases (macOS, Windows, Linux)

Hector uses a unified configuration system. You can start with CLI flags (which seed the config) or a config file directly.

go install github.com/verikod/hector/cmd/hector@latest export OPENAI_API_KEY="sk-..." hector serve --model gpt-4o --tools --studio

RAG in one command (with MCP parsing optional):

hector serve \ --model gpt-4o \ --docs-folder ./documents \ --mcp-url http://localhost:8000/mcp \ --mcp-parser-tool convert_document_into_docling_document

cat > config.yaml <<'EOF' version: "2" llms:  default:  provider: openai  model: gpt-4o  api_key: ${OPENAI_API_KEY} agents:  assistant:  llm: default  tools: [search] server:  port: 8080 EOF hector serve --config config.yaml --studio

• Unified Configuration: CLI flags seed a YAML config file for repeatability. JSON Schema available via hector schema.
• Programmatic API: Build agents in Go (pkg/api.go), including sub-agents and agent-as-tool patterns.
• RAG: Folder-based document stores, embedded vector search (chromem), native PDF/DOCX/XLSX parsers, optional MCP parsing (Docling).
• Vector DBs: Embedded chromem (default), or external (Qdrant, Pinecone, Weaviate, Milvus, Chroma).
• Persistence: Tasks and sessions can use in-memory or SQL backends (sqlite/postgres/mysql via DSN).
• Observability: Metrics endpoint and OTLP tracing options.
• Checkpointing: Optional checkpoint/recovery strategies.
• Auth: JWT/JWKS support at the server layer.
• Guardrails: Input validation, prompt injection detection, PII redaction, and tool authorization.
• A2A-native: Uses a2a-go types and JSON-RPC/gRPC endpoints.

• Getting Started
• Configuration Guide
• Guardrails Guide
• RAG Guide
• Tools Guide
• Core Concepts
• Blog & Tutorials

AGPL-3.0 (see LICENSE).