parlant
Python framework for building deterministic, policy-guided customer chat agents powered by LLMs.
Deploy rule-bound client chat agents that run on your infrastructure and follow explicit policies.
Parlant is an open-source Python platform designed for building customer-facing chat agents that adhere to operational policies. It suits technical teams in engineering and architecture firms who need reliable client intake, project status querying, or technical support bots without handing full control over to unconstrained models.
What it is
Parlant lets you define agent behavior using plain-language policies, canned responses, and custom Python tools. It evaluates context during live conversations to apply relevant rules, validate tool calls before execution, and review outbound messages before sending them to clients.
Who it's for
It suits software developers and IT teams who want full programmatic control over customer interactions, custom backend integrations, and LLM providers. It is not suitable for non-technical users looking for a visual no-code builder.
Key features
Policy-based orchestration
Dynamically applies plain-language business rules as customer interaction context evolves.
Tool and message reviews
Validates system calls and outgoing responses before execution to prevent policy breaches.
Adaptive effort spending
Allocates extra processing power and tokens only during high-stakes conversational turns.
Embedded chat and handoff
Provides a ready-to-use web chat widget with built-in human escalation capabilities.
Persona simulation testing
Runs automated conversational tests against defined customer personas before pushing to production.
Up and running in 3 steps
Our honest verdict
What's great
- Performs pre-execution checks on tool calls and outbound messages to ensure compliance.
- Includes built-in testing tools to simulate customer interactions against target personas.
- Directly connects to major model providers like OpenAI, Anthropic, Gemini, and LiteLLM.
- Includes a front-end chat widget and human handoff workflow out of the box.
Where it falls short
- Requires ongoing API expenditures with LLM providers based on usage volume.
- Demands Python development expertise to define tools, set up servers, and write policies.
- Requires managing your own server infrastructure unless using their managed hosting option.
Pricing & license
Free and open-source under the Apache-2.0 license. Operating costs include model API fees from your chosen LLM provider and host server infrastructure, or optional managed hosting on the Parlant Dashboard.
Open remote AI-training roles
768 open remote AI-training roles, updated November 14, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
How does Parlant control AI output?
It evaluates conversation state against your plain-language policies and validates tool calls and messages before they are processed.
What model providers does Parlant support?
It connects directly with OpenRouter, OpenAI, Anthropic, Gemini, LiteLLM, and custom providers.
