DOCSEARCHbyAlgolia
Submit Docs

AgentWeave Developer Documentation

website·Indexed 30 days ago

https://developer.agentweave.ai

An AI collaboration platform for creating and managing AI agents, agent teams, and virtual chatrooms. Features include RAG template configurations, A2A Protocol for agent-to-agent communication, Model Context Protocol (MCP) integration, and proxy support for LLM providers like OpenAI, Claude, and Gemini. Documentation covers local deployment via Docker, REST API usage, and the creation of AI-generated web applications.

Tokens
16.1K
Snippets
24
Records
117
Agent score
50%

What's inside AgentWeave

  1. Overview of AgentWeave A2A Protocol implementation

    AgentWeave implements the Agent2Agent (A2A) Protocol version 0.3, which is compatible with the @a2a-js/sdk 0.3.x JSON-RPC schema. A2A is an open-source protocol designed for standardized communication and collaboration between different AI agents across platforms and vendors.

  2. Overview of AgentWeave Local Deployment

    AgentWeave Local Deployment allows users to run the entire platform (backend, agent runtime, database, object storage, and login) on their own hardware using Docker. This ensures that data, models, and compute remain under the user's control.

    Core use cases include:

    1. Private Agent Execution: Run agents using any OpenAI-compatible model provider entirely within a local network.
    2. Web-redirect Agents: Create shortcuts in the local portal that redirect to agents hosted on external sites.
    3. Cloud-to-Local Bridge: Use a secure edge tunnel to bind a cloud-based AgentWeave agent to local compute, allowing public users to interact with an agent served by local hardware.

  3. Introduction to AgentWeave AI Platform

    AgentWeave is an AI collaboration platform that enables developers to create, manage, and deploy intelligent AI agents. It provides a set of tools and features designed to streamline the building of agents for various operational scenarios.

  4. Overview of AgentWeave capabilities

    AgentWeave is a lightweight intelligent conversation platform designed to help developers integrate AI agents into applications and websites. It provides three primary integration methods:

    1. REST API: Enables applications to access intelligent agents for conversational interactions via backend calls.
    2. Chat Widget: A plug-and-play web component that can be embedded into any website to provide a user-facing chat interface without extensive development.
    3. MCP Server Interface: Every agent automatically exposes a Model Context Protocol (MCP) endpoint. This allows other AI systems or tools to programmatically discover, query, and consume the agent's capabilities, facilitating AI-to-AI interoperability.

  5. Integrate external tools with AI Teams

    AI Teams support integration with external tools to bridge the gap between planning and execution in real-world workflows. For example, an AI Team can generate travel recommendations and then integrate with a map tool to display actual routes, enhancing decision-making efficiency and user experience.

  6. Understand the Model Context Protocol (MCP) architecture

    The Model Context Protocol (MCP) is an open standard designed to connect LLM applications with external data sources and tools. It operates on a client-server architecture:

    • MCP Servers: Lightweight programs that expose specific capabilities and data sources (local or remote) via the standardized protocol.
    • Host Applications (MCP Clients): AI tools, chatbots, or IDEs (e.g., Claude Desktop, Cursor, Windsurf, Cline, Q Developer) that maintain one-to-one connections with MCP servers to retrieve context and improve model output accuracy.

  7. Create an A2A Team for RAG-based collaboration

    The A2A Team template uses a Retrieval-Augmented Generation (RAG) architecture for multi-agent collaborative retrieval and document parsing.

    Adding Agents to an A2A Team:

    • Via Official Library: Navigate to "Functional Agents" $\rightarrow$ "+ Select" $\rightarrow$ Browse Official Library/Team Shared Agents $\rightarrow$ Select agents (e.g., Mistral, DeepSeek) $\rightarrow$ Confirm.
    • Via Input Link: Use "Add via Input" and provide the agent URL.

    URL Format Specification: http://{Domain}/.well-known/agent.json

    • Domain: The deployment address (e.g., rest.agentweave.ai).
    • agentId: The unique identifier of the target agent.
    • your_token (Optional): Authentication token (passing via header is recommended for security).

  8. Common use cases for MCP servers

    The Model Context Protocol (MCP) is typically used for three primary patterns:

    1. Data Queries: Fetching external information such as weather data, executing database queries, or performing search engine inquiries.
    2. Tool Execution: Triggering external actions like calling scripts, invoking APIs, or running automated processes.
    3. Cross-agent Collaboration: Facilitating the exchange of data and state between different AI agents or services.

  9. Retrieve an Agent Card for A2A discovery

    The Agent Card provides metadata about an agent, including its capabilities, supported input/output modes, skills, and the A2A protocol endpoint URL. The card is available at two well-known paths for any given agent ID.
    GET https://rest.agentweave.ai/api/v1/agents/{agentId}/.well-known/agent-card.json
OR
GET https://rest.agentweave.ai/api/v1/agents/{agentId}/.well-known/agent.json

  10. Publish and test AI agents

    Publishing an agent involves an automated review process to ensure compliance with platform standards.

    Publication Process:

    1. Submit: Click the 'Publish' button to submit the agent for review.
    2. Review: The agent remains private while the platform reviews it.
    3. Activation: Once passed, the agent is activated. If it fails, you will receive feedback for modifications.

    Post-Publication:

    • Test: Use the 'Try It' button to verify the agent works as expected.
    • Manage: Use the 'Manage' button to enter the configuration and monitoring interface.

  11. Upgrade or cancel AgentWeave pricing plans

    AgentWeave offers several tiers: Personal ($10/mo), Pro ($20/mo), Team ($220/mo), Team Plus ($550/mo), and Enterprise (custom/private deployment). New users receive a 14-day free trial of the Personal plan.

    To Upgrade:

    1. Log in and navigate to the 'Pricing' page in the top navigation bar.
    2. Select the desired plan and click 'Subscribe' or 'Contact Us'.
    3. The upgrade takes effect immediately; check 'Settings' for the next billing date.

    To Cancel:

    1. Go to the 'Settings' page via the user avatar in the top right corner.
    2. Locate the current subscription and click 'Cancel Subscription'.
    3. Automatic charging stops for the next cycle, but features remain active until the end of the current billing period.

  12. Cancel an AgentWeave Organization subscription plan

    Only the organization's Owner can cancel an organization subscription. Log in to the AgentWeave Official Website, click the 'Organization' tab at the top of the Home page to go to the 'My Organizations' page, and select the target organization. Navigate to the 'Team Plan' tab, locate the 'Active' plan, and click the 'Cancel Subscription' button.