Why Parlant

Welcome to Parlant.

Parlant is an open platform for building customer-facing chat agents with AI. You describe how the agent should behave, and Claude CodeCodexCursorLovableBase44 implements it for you.

Parlant provides the rest: an engine that keeps the agent on your rules in every conversation, plus the parts every customer-facing agent needs, from a chat widget and human handoff to testing, hosting, and analytics.

Because the agent runs on your own code, it can use any system you can reach and any model you choose, and—unlike closed CX platforms—you pay model providers directly instead of an overpriced fee per resolution.

What you'll learn

• What building and changing an agent with Parlant looks like day to day
• Where to start, from installation to deployment
• Why customer-facing agents are harder to build from scratch compared to AI assistants
• How Parlant's features ramp up a productino-ready agent out of the box, and when you'd need each part

The quickstart guide takes you from installation to a working test agent, connected to a frontend chat widget.

What building with Parlant feels like

Imagine you're working on a lead-generation agent for your company's website, and a new feature launches in 1 hour.

• Monday, 8:43 Marketing sends you the final FAQ for the new Pro plan, which launches at 10:00.
• 8:45 You tell your local coding agent: "Make it answer Pro questions from this FAQ. If a Basic customer hits their project limit, mention Pro once. Never pitch it to Pro customers." Then you go make a coffee.
• 9:00 Your coding agent adds a policy and uses each customer's plan to decide when Pro is relevant.
• 9:01 You tell your coding agent: "Test it with two simulated personas: a Basic customer at their project limit, and a Pro customer." While they run, you grab a cookie.
• 9:10 You open the two test sessions on the Dashboard and check that they went as you expected: the Basic customer hears about Pro once, and the Pro customer never does.
• 9:15 You push the change. At 10:00, your agent is ready for launch-day questions.

The lead-generation agent is ready for launch day, with no vendor ticket, no fiddling in deep menus, and no engineering sprint.

Why it actually works

Creating AI agents isn't hard by itself. The hard part is making sure they do their job correctly and consistently: even with a customer who mixes different topics, or pushes back against your policies.

That's where Parlant's got your back, because customer-facing agents face pressures that assistant-style agents never do:

	AI Assistant	Customer-Facing Agent
Who uses it	A cooperative, informed user	A possibly confused, impatient, or hostile customer
User expectation	The agent must do as it's told	The agent should proactively know what to do and how to help
Cost of a mistake	A retry or a revised draft	Financial losses, brand damage, and customer relationships
Success criteria	Satisfy each user's own goals	Satisfy users, but never violate business integrity
Supervision	The user corrects it iteratively	Nobody is there to catch mistakes during the interaction
Structure	Conversation is led by the user	The agent must often stick to specific operating procedures
Language	Free to improvise	On-policy at all times, sometimes with exact wording
Latency	Proportional to the task at hand	Always upper-bounded at a conversational pace
Accountability	Nice to have on occasion	Behavior must always be explainable

Most AI products, and the code a coding agent writes on top of them, are built for that left column: for assistant agents.

But Parlant is built for the right column, and here are its unique features that do the work:

Feature	What it is	What's the value
LLM integration	Integrates any model from any provider, paid at the provider's market price with no markup.	You want the best model for the price, and the freedom to switch when a better one comes out.
Agentic orchestration	Brings in the relevant policies as the customer's needs change, without predefined conversation paths.	Customers mix topics, change their minds, or come back to earlier requests.
Context discovery	Adds relevant policies as topics emerge, and keeps them in play as the conversation wanders.	Your agent has more rules than fit comfortably in one prompt.
Cross-session memory	Connects Parlant to mem0 so your agent remembers customer facts across chats.	Returning customers shouldn't have to repeat themselves.
Session compaction	Summarizes long chats, guided by your policies, keeping commitments and tool results intact.	Conversations run long, and token costs or context limits start to bite.
Conversation API	One persistent history of messages, replies, actions, and status, shared by the agent and human operators.	You're building your own frontend or connecting other channels.
Chat UI	A ready-made chat widget with streaming, returning guests, sign-in, and rich results.	You want a production chat interface without building one.
Authentication	Separates signed-in customers, guests, and admins, and enforces who owns each conversation.	Real customers, and their data, are about to use your agent.
Content moderation	Checks customer messages and agent replies with the moderation models you configure.	Your agent is public, and abuse or unsafe output is a real risk.
Adaptive effort	Raises the effort spent on a turn when your rules recognize a high-stakes moment.	Some moments, like disputes or safety issues, deserve more care than others.
Policy focus	Reminds the model of the policies that apply to each reply.	Details in long policies are getting overlooked.
Tool validation	Checks proposed tool calls against your instructions before they run.	Tools have real consequences, like refunds, cancellations, or account changes.
Message validation	Checks replies against your instructions, approved wording, and tool results before delivery.	A wrong promise or an off-policy reply would be costly.
Canned responses	Approved wording examples that your agent adapts to each conversation.	You want your agent to speak in your tone of voice and with your wording choices, from everyday phrasing to the exact wording of a disclosure.
Rich input & output	Carries attachments and UI payloads alongside the conversation.	Customers send files, or you want cards and structured results in the chat.
Steering guidelines	Short policies that adjust how your agent handles a kind of moment across every topic, such as frustration or confusion.	You want consistent behavior everywhere, not per flow.
Personalization variables	Resolves facts about the current customer, with group and shared fallbacks.	Answers depend on who the customer is, such as their plan, account, or region.
Conversation analytics	Labels conversations when a situation or policy applies, ready to query.	You need to know what customers ask about and how conversations end.
Human handoff	Lets a person take over the same chat, then hand it back to your agent.	Some conversations need a human, and nothing should be lost in the switch.
Spend management	Tracks conversation costs and enforces session budgets.	You need predictable costs at scale.
Agent observability	Links each reply to the policies, tool calls, and checks behind it.	You need to explain or debug why your agent said something.
Managed hosting	The Dashboard hosts your agents, with conversation storage, TLS, and scaling.	You'd rather not run the infrastructure yourself.

Next steps

1. Install & run: install Parlant, create a project, and talk to your first agent.
2. Take your agent live: deploy on the Dashboard, connect your app, and set up real users.
3. Integrate it into your app or site: see the chat widget running against your agent, then put it in your app.