# IMFA Agents > IMFA Agents is a Morocco-built platform where an AI builder creates customizable AI agents, chatbots, and agentic workflows from a conversation. Describe the agent you need in plain language, then teach, test, publish, embed, call, or connect it to your application. Canonical site: https://www.imfa.app/ Created by IMFA Solutions in Morocco: https://www.imfa.solutions/ Full product documentation: https://www.imfa.app/llms-full.txt Agent integration reference: https://www.imfa.app/agents.md ## Product overview IMFA Agents is a web platform for building customer-facing support agents, sales agents, assistants, chatbots, and connected agentic workflows from a conversation. The AI builder converts a natural-language brief into an agent identity, system prompt, personality, greeting, suggested questions, skills, model configuration, knowledge setup, and widget design. Owners can refine the draft in chat, preview it live, and publish only when ready. The core idea is agents building agents. The builder agent does the configuration work, while the user stays in control of consequential actions such as publishing, changing integrations, spending credits, and exposing an agent publicly. ## Main capabilities - Conversational creation and editing of AI agents. - Draft, published, and paused agent states. - Live website widget preview and one-line embed script. - Public share page and QR code for each published agent. - Knowledge from PDFs, Word documents, spreadsheets, plain text, and public webpages. - Friendly, professional, or playful personalities. - Optional skills for human writing, multilingual replies, Moroccan Darija, discovery questions, empathy, persuasion, and closing. - Custom system prompts, greetings, suggested questions, avatars, colors, fonts, message styles, corner radius, shadows, bubble size, and left-to-right or right-to-left direction. - Gateway-backed models from OpenAI, Anthropic, Google, Kimi, and DeepSeek without per-agent provider keys. - Website, Telegram, WhatsApp, and Discord customer channels. - Live voice chat and public calling in beta for eligible Pro accounts. - Lead capture for names, email addresses, phone numbers, and conversation summaries. - Lead inbox, CSV export, chat-session metrics, message metrics, lead metrics, and per-agent credit usage. - Account-scoped MCP access for Codex, Claude, Cursor, VS Code, Zed, and compatible clients. - Agent-scoped REST API keys for server-side integrations. - Built-in prompt-injection defenses that keep system instructions private and treat retrieved knowledge as untrusted data. - Standard and Pro plans, monthly credits, optional non-expiring credit packs, and billing management. - Public product changelog and RSS feed. ## Typical workflow 1. Describe the desired agent in a sentence. 2. Let the builder agent create a configured draft. 3. Add business instructions, knowledge, personality, skills, model settings, channels, and visual styling. 4. Test the conversation in the live preview or voice interface. 5. Publish explicitly when the draft is ready. 6. Install the website script, share the public link, connect messaging channels, use MCP, or call the REST API. 7. Review sessions, captured leads, credit use, and performance, then update and republish the draft. ## MCP server Endpoint: https://www.imfa.app/api/mcp Authentication: account-scoped bearer token created at https://www.imfa.app/app/mcps Transport: MCP over HTTP using POST, GET, and DELETE. Tokens are shown once, stored only as hashes, scoped, expiring, and independently revocable. Read-only tokens expose safe reads. Full access tokens can also create, update, publish, and chat. Available tools: - agents_list: list owned agents and resolve exact slugs. - agents_get: inspect normalized draft and published chatbot/v1 theme state. - agents_api_docs_get: return the complete REST API integration guide without generating or revealing API keys. - create_agent: create a configured, optionally themed, publish-ready draft without publishing it. - update_agent: update identity, prompt, personality, greeting, or server-side theme draft. - publish_agent: publish the current draft only after an explicit user request. - agents_integration_get: return safe embed code and REST endpoint URLs without secrets. - agents_chat: send a message to a published agent with persisted session context and credit usage. - agents_sessions_list: list recent REST and MCP chat sessions. - agents_session_messages_list: read conversational text from a selected session. Detailed MCP setup, client examples, tool arguments, and operating rules are in https://www.imfa.app/agents.md and https://www.imfa.app/llms-full.txt ## REST API Base URL: https://www.imfa.app/api/v1 Authentication: agent-scoped bearer key created inside the agent API view or at https://www.imfa.app/app/api-keys REST keys are for trusted server code. They are shown once and stored only as hashes. The API is intentionally not a public browser CORS API because requests spend the owner's credits. Endpoints: - GET /api/v1/agents - POST /api/v1/agents/{slug}/chat - GET /api/v1/agents/{slug}/sessions - GET /api/v1/agents/{slug}/sessions/{sessionId}/messages - GET /api/v1/agents/{slug}/leads The chat body accepts message and an optional sessionId. Reuse the returned sessionId to continue context. List endpoints use cursor pagination with a default of 20 items and a limit from 1 to 100. ## Public resources - Home: https://www.imfa.app/ - French home: https://www.imfa.app/fr - Changelog: https://www.imfa.app/changelog - Changelog RSS: https://www.imfa.app/changelog.xml - French changelog: https://www.imfa.app/fr/changelog - French changelog RSS: https://www.imfa.app/fr/changelog.xml - Contact: https://www.imfa.app/contact - French contact: https://www.imfa.app/fr/contact - Privacy policy: https://www.imfa.app/legal/privacy - French privacy policy: https://www.imfa.app/fr/legal/privacy - Terms of service: https://www.imfa.app/legal/terms - French terms of service: https://www.imfa.app/fr/legal/terms - Full LLM documentation: https://www.imfa.app/llms-full.txt - Agent and developer reference: https://www.imfa.app/agents.md - Sitemap: https://www.imfa.app/sitemap.xml English is the default public language. French equivalents use the /fr URL prefix. Canonical and reciprocal hreflang metadata identify every translated pair. ## Access notes - The dashboard, MCP token manager, API key manager, builder, billing, settings, and private documentation require an authenticated IMFA Agents account. - MCP create, update, publish, and chat tools must run only when the user asks for that action. - Creating or updating an agent changes a draft. It does not publish unless publish_agent is called after explicit approval. - MCP and REST chat calls persist sessions and spend the owner's credits. - Voice and public calling are beta features and require an eligible Pro account with available credits. - Website knowledge and custom model controls are Pro features.