bolna-graph-agents

Installation
SKILL.md

Bolna Graph Agents

Build structured, multi-step voice conversations as a directed graph of nodes and edges. Each node has one job and one set of transitions. Routing is deterministic where you want it (expression edges, events) and LLM-driven where you need flexibility.

When to use a graph agent

Use graph agent Use simple_llm_agent
Distinct stages with different objectives (greet → qualify → collect → confirm → close) A single objective like answering FAQ
Deterministic branches (working hours, retry counts, language) Free-flowing chat where any topic might come up
External events drive the conversation (payment confirmed, link clicked) Pure speech-driven flow
Compliance checkpoints that must not be skipped Tone-only personality differences
High-volume flows where static nodes save latency and LLM cost Low-volume prototypes

If you can describe the call as a flowchart on a whiteboard, a graph agent fits. If it's "one persona, many possible topics," stay with a prompt agent.

Core concepts in one minute

Installs
55
Repository
bolna-ai/skills
GitHub Stars
3
First Seen
May 20, 2026
bolna-graph-agents — bolna-ai/skills