DOCSEARCHbyAlgolia
Submit Docs

DesireCore Documentation

website·Indexed 30 days ago

https://docs.desirecore.com

A delegative interaction Agent operating system designed for cultivating and hosting digital companions. The platform features a local-first architecture, AgentFS for agent identity and runtime policy configuration via agent.json, and a version management system for team-based agent development. Documentation covers agent types, automation tools like Heartbeat, and version history from v10.0.0 through v10.1.0.

Tokens
307K
Snippets
459
Records
3K
Agent score
54%

What's inside DesireCore

  1. Overview of Web Access v2 capabilities

    Web Access v2 allows agents to use a controlled browser to interact with web pages. It is specifically designed for scenarios where standard web scraping fails, such as dynamic pages, authenticated backends, complex forms, Single Page Applications (SPAs), internal systems, and pages requiring screenshot verification.

  2. Overview of the DesireCore Agent Market

    The Agent Market is the central hub for discovering and acquiring agents, skills, and applications. It allows users to browse official and community-provided entries and install them locally. The market is accessed via the 'Market' icon in the main navigation bar and is organized into three primary tabs:

    • Agent: Agent templates that become personal assistants upon installation.
    • Skill: Skill packs used to extend the capabilities of agents.
    • 应用 (Apps): Third-party applications or service entry points integrated with DesireCore.

  3. Overview of Scheduled Tasks

    Scheduled tasks allow an agent to execute a specific instruction at a future time or on a fixed cycle. They are ideal for reminders, periodic summaries, automated reports, and repetitive background work. Unlike 'Heartbeat' (continuous monitoring) or 'Sleep' (short pauses within a current task chain), scheduled tasks trigger a new execution session in the background and do not inherit the temporary context of the current input box.

  4. Overview of Scheduled Tasks in DesireCore

    Scheduled tasks allow an agent to execute a specific prompt at a designated time or on a repeating schedule. They run independently in the background and are ideal for reminders, periodic summaries, and automated reports.

    Key Distinction: Unlike Sleep (which pauses a current task chain) or Heartbeat Monitoring (which notifies only when a condition changes), a scheduled task starts a completely new run. Because the scheduled run does not inherit temporary context from the conversation where it was created, the prompt must be self-contained.

  5. Overview of DesireCore Intelligent Task Orchestration Engine

    The DesireCore Intelligent Task Orchestration Engine acts as a "Project Manager" that converts a single natural language requirement into a coordinated delivery involving multiple agents. It automates the process of intent recognition, task decomposition, and agent assignment, handling the entire lifecycle from requirement to final delivery.

  6. Overview of DesireCore AI Provider Configuration

    DesireCore supports a wide range of AI service types, including conversation, reasoning, tool calling, vision understanding, embeddings, speech synthesis (TTS), speech recognition (ASR), image generation, video generation, music generation, OCR, and Computer Use.

    Note that generation-specific models (e.g., image_gen, video_gen, music_gen, tts, asr) do not appear as standard chat models and are instead invoked by specific tools or skills. The list of available models is updated via the config-center, and actual availability depends on the client, account permissions, and remote configuration. For real-time pricing and context lengths, refer to the 'Settings > Compute Services' (设置 > 算力服务) menu in the application.

  7. Overview of DesireCore Core Features

    DesireCore provides several key features for AI interaction and automation:

    • Conversations: Natural chat with an AI companion.
    • Teaching System: Ability to define rules and preferences for the companion via chat.
    • Delegation: Allows the companion to plan and execute tasks autonomously.
    • Super Document: A system for reviewing AI-generated documents similarly to a code review.
    • Memory System: A persistent memory that allows the companion to remember preferences and improve over time.
    • Heartbeat Monitor: A proactive system where the companion performs periodic checks and notifies the user.

  8. Overview of DesireCore Agent OS

    DesireCore is a delegation-based interactive Agent OS designed to move beyond traditional Q&A AI assistants. Its core operational philosophy is based on a cycle of: Teaching the agent $\rightarrow$ Agent planning $\rightarrow$ Agent execution $\rightarrow$ Agent reporting $\rightarrow$ Agent evolution. This allows users to teach AI agents their specific way of working so the agents can act as capable assistants.

  9. Overview of DesireCore Six Primitives

    All interactions in DesireCore are based on six fundamental primitives that govern how users collaborate with agents. These primitives form a continuous optimization cycle (Teach → Demonstrate → Plan → Execute → Reflect → Clarify) to increase agent reliability over time.

  10. Overview of DesireCore Core Concepts

    DesireCore is built around the core philosophy of 'cultivating and hosting digital companions.' Rather than treating AI as a simple tool, it positions agents as 'digital colleagues.' The system is designed to move from a 'question-and-answer' interaction to a 'delegation' model ('I teach you how to do it'), ensuring that agents are flexible like humans but controllable like software.

  11. Overview of Model Context Protocol (MCP)

    The Model Context Protocol (MCP) is an open standard that acts as a "universal socket" between AI agents and external services (such as GitHub, Slack, databases, or custom APIs). It allows agents to securely access external tools and data sources without requiring specialized integration code for every individual service.

  12. Overview of DesireCore Integration Architecture

    DesireCore uses a multi-layered integration architecture to connect with external services and devices. The architecture consists of four primary complementary components:

    • Built-in Tools: Provide foundational capabilities such as file operations, search, command execution, and network requests.
    • MCP (Model Context Protocol): An extension interface used to connect to any external service that provides an API.
    • Skill Packs: Templates that orchestrate multiple tools into professional, repeatable workflows.
    • Computer Use: A fallback mechanism for controlling GUI applications that lack APIs, utilizing HostAgents or CUA Drivers.