Prompts

Create versioned prompt templates, resolve Liquid variables, and apply prompts to agents

The Prompts module manages your workspace prompt library. Templates support Liquid variables, version history, and agent linking. MCP clients discover the same prompts via prompts/list.

Setup

import { Chanl } from '@chanl/sdk';

const chanl = new Chanl({
  apiKey: process.env.CHANL_API_KEY,
});

List and get

const { data: prompts } = await chanl.prompts.list({
  status: 'active',
  category: 'support',
});

const { data: prompt } = await chanl.prompts.get(promptId);

Create and update

const { data: created } = await chanl.prompts.create({
  name: 'Support Greeting',
  content: 'Hello {{customer_name}}! How can I help?',
  category: 'support',
  variables: [
    {
      key: 'customer_name',
      required: true,
    },
  ],
});

// PATCH — content changes create a new version
await chanl.prompts.update(created.id, {
  content: 'Hi {{customer_name}}, thanks for reaching out.',
});

Versions and revert

const { data: versions } = await chanl.prompts.getVersions(promptId);

await chanl.prompts.revert(promptId, { version: 2 });

Resolve variables

const { data } = await chanl.prompts.resolve(promptId, {
  variables: {
    customer_name: 'Alice',
  },
});

console.log(data.content);
console.log(data.unresolvedVariables ?? []);

Apply to an agent

await chanl.prompts.apply(promptId, {
  agentId: '507f1f77bcf86cd799439011',
  syncToPlatform: true,
});

Python

import asyncio
from chanl import AsyncChan

async def main():
    async with AsyncChan(api_key="chanl_key_...") as client:
        prompts = await client.prompts.list(status="active")
        resolved = await client.prompts.resolve(
            prompt_id,
            variables={"customer_name": "Alice"},
        )
        print(resolved)

asyncio.run(main())

CLI

chanl prompts list
chanl prompts create -f prompt.json
chanl prompts revert <id> --to-version 2
chanl prompts resolve <id> -f vars.json
chanl prompts apply <id> --agent <agent-id>

See CLI reference and MCP prompts.

On this page