awesome-repositories.com
Blog
MCP
awesome-repositories.com

Discover the best open-source repositories with AI-powered search.

ExploreCurated searchesOpen-source alternativesSelf-hosted softwareBlogSitemap
ProjectMCP serverAboutHow we rankPress
LegalPrivacyTerms
© 2026 Bringes Technology SRL·VAT RO45896025·hello@awesome-repositories.com
google-antigravity avatar

google-antigravity/antigravity-sdk-python

0
View on GitHub↗
3,966 stars·1,558 forks·Python·Apache-2.0·0 viewsantigravity.google/product/antigravity-sdk↗

Antigravity Sdk Python

This is a Python SDK for building asynchronous AI agents. A single async session connects to an agent backend, sends prompts (text or multimodal content such as images and documents), and streams results back as steps, text tokens, model thoughts, and tool calls. The session handles backend startup, authentication, tool wiring, hook registration, and cleanup, so developers don't write the agentic loop by hand. Agents run through a locally running harness on the developer's own machine, and the package ships with a pre-compiled platform-specific binary.

What sets it apart is how much of the agent's behavior can be controlled from outside the agent itself. A declarative policy engine applies allow, deny, and ask-user rules to tool calls, with exact tool names taking priority over wildcards and optional argument-based conditions. Tools can also be hidden from the model entirely. Layered lifecycle hooks work like middleware around sessions, turns, and operations, covering logging, guardrails, rate limiting, and error recovery, with scoped state shared across those stages. Event-driven triggers run as background tasks on intervals, on file changes, or on custom external events, and push messages into a live session so the agent can act without a user prompt.

Tooling and collaboration cover the rest of the surface. Ordinary Python functions can be registered as tools, and external Model Context Protocol servers can be connected alongside them. Several agents can share one chat room, either in synchronized rounds or asynchronously with reactive wake-ups, and an image-generating agent can be chained with a reviewing agent. Agents can also run against locally hosted models entirely offline. Supporting features include session history inspection and size limits, token usage tracking, optional telemetry export, configurable model retries, reusable skill packages, and enterprise platform sign-in.

The repository includes guides and small single-file examples, each teaching one concept such as agents, streaming, tools, policies, hooks, and structured output, plus a full interactive command-line agent that asks for approval before sensitive actions.

Features

  • Agent SDKs - Offers an asynchronous Python SDK that connects to an agent backend, sends text or multimodal prompts, and streams steps, tokens, thoughts, and tool calls.
  • Autonomous AI Agent Frameworks - Provides an async Python SDK for building agents that connect to a backend, call tools, and hold multi-turn conversations without a hand-written agentic loop.
  • AI Agent Development Tools - Builds asynchronous AI agents by wiring tools, hooks, and default policies behind one async session, so developers skip hand-writing the agentic loop.
  • Event-Driven Agent Loops - Runs background triggers on intervals, file changes, or external events that push messages into a live session so agents act without a user prompt.

AI search

Explore more awesome repositories

Describe what you need in plain English — the AI ranks thousands of curated open-source projects by relevance.

Start searching with AI
  • Agent Lifecycle Hooks - Stacks lifecycle hooks around sessions, turns, and operations like middleware to add rate limiting, audit logging, and error recovery without changing agent logic.
  • Agent Harnesses - Connects to a locally running agent harness shipped as a pre-compiled binary, so agents execute directly on the developer's machine.
  • Tool Allow and Deny Patterns - Applies declarative allow, deny, and ask-user rules to agent tool calls, with exact names outranking wildcards, so only permitted actions run.
  • Agent Governance - Governs what an autonomous agent may do by gating tool calls with allow, deny, and approval rules plus lifecycle hooks for guardrails.
  • Event-Triggered Agent Runs - Runs background triggers on intervals, file changes, or custom external events that push messages into a live session so the agent acts without a prompt.
  • Asynchronous Agent Task Triggers - Runs event triggers as long-lived background tasks that push messages into a live agent session so it can act without a user prompt.
  • Runtime Event Interception - Runs custom code at set stages of agent execution so developers can observe, gate, or change behavior for logging, guardrails, and cleanup.
  • Agent Action Policies - Applies declarative allow, deny, and ask-user policies to agent tool calls, with a read-only default that can be widened to allow write tools.
  • Layered Tool Permission Gates - Checks each tool call against allow, deny, and ask-user rules before it runs, with exact names outranking wildcards and argument conditions.
  • Middleware Lifecycle Hooks - Stacks lifecycle hooks around sessions, turns, and operations like middleware, enabling logging, guardrails, rate limiting, and error recovery without touching agent logic.
  • Agent Backend Managers - Establishes a live agent backend session, handling process startup, transport setup, authentication, and health checks inside an async context manager.
  • Custom Function Registrations - Registers ordinary Python functions as callable tools the agent can invoke to fetch information or take actions during a conversation.
  • Agent Prompt Submission - Sends text or multimodal prompts to a connected agent and streams completed steps back asynchronously, with turn cancellation and idle checks.
  • Agent Tool Server Connections - Connects agents to external Model Context Protocol servers so they can discover and call outside tools alongside built-in and custom ones.
  • Custom Tool and Reward Definitions - Registers ordinary Python functions, sync or async, as agent tools and runs them on request without blocking the event loop.
  • Multi-Agent Collaboration Systems - Coordinates several agents in one chat room, in synchronized rounds or asynchronously, and chains image generation with a reviewing agent.
  • Model Context Protocol - Gives agents access to custom Python functions and external Model Context Protocol servers so they can fetch information and take actions.
  • MCP Server Connections - Connects external MCP servers and makes their tools available to the agent alongside its built-in capabilities.
  • Agentic Toolkits - Registers ordinary Python functions as agent tools and connects external Model Context Protocol servers alongside them to extend what agents can do.
  • Multi-Agent Chat Rooms - Runs several agents in a single chat room, either in synchronized parallel rounds or asynchronously with reactive wake-ups when new messages arrive.
  • Action Approval Gates - Pauses a tool call and asks a human-supplied handler to approve or deny it, failing fast at setup if no handler exists.
  • Async Agent Session Lifecycles - Wraps the whole agent session in an async context manager that starts the backend, wires tools and hooks, and cleans up automatically on exit.
  • Console Agent Output Streams - Streams conversational text tokens as the agent produces them, giving responsive console output without waiting for the full response.
  • Local Model Inference - Runs agent workflows entirely offline with locally hosted open models, finishing tasks without network access to a hosted model service.
  • Offline Model Runners - Runs agent workflows against locally hosted models entirely offline on the user's own hardware without sending data to external services.
  • Real-Time Tool Call Streams - Exposes model thoughts and tool call requests through dedicated async streams so applications can watch or intercept them in real time.
  • Shared Room Agent Orchestration - Runs several agents in a single chat room, in synchronized rounds with parallel turns or asynchronously with reactive wake-ups.
  • Multimodal Agent Capabilities - Passes images, videos, audio, and documents to the agent alongside text prompts using in-memory content objects or file paths with automatic type detection.
  • AI Agent Session Managers - Manages the full agent session lifecycle behind one async context manager, including binary discovery, tool wiring, hook registration, and safe default policies.
  • Conversation Step Streaming - Yields agent steps, text tokens, thoughts, and tool calls through async iterators so interfaces can render output while the model is still generating.
  • Event-Driven Triggers - Runs background triggers on schedules or outside events that push messages into a live agent session so it acts without a user prompt.
  • Action Policy Enforcement - Checks each operation before it runs and returns an allow or deny decision, optionally rewriting tool arguments first.
  • AI Prompt Sending APIs - Sends text or multimodal prompts such as images and documents to an agent and returns the complete response once finished.
  • Token Streaming - Streams tokens, model thoughts, and tool calls as incremental chunks so interfaces can display output while the model is still generating.
  • Tool Access Controls - Declares allow, deny, and ask-user rules for tool calls through simple builder functions, with exact tool names taking priority over wildcards.

    • Star history

    Star history chart for google-antigravity/antigravity-sdk-pythonStar history chart for google-antigravity/antigravity-sdk-python

    How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.

    Frequently asked questions

    What does google-antigravity/antigravity-sdk-python do?

    This is a Python SDK for building asynchronous AI agents. A single async session connects to an agent backend, sends prompts (text or multimodal content such as images and documents), and streams results back as steps, text tokens, model thoughts, and tool calls. The session handles backend startup, authentication, tool wiring, hook registration, and cleanup, so developers don't write the agentic loop by hand. Agents run through a locally running harness on the developer's…

    What are the main features of google-antigravity/antigravity-sdk-python?

    The main features of google-antigravity/antigravity-sdk-python are: Agent SDKs, Autonomous AI Agent Frameworks, AI Agent Development Tools, Event-Driven Agent Loops, Agent Lifecycle Hooks, Agent Harnesses, Tool Allow and Deny Patterns, Agent Governance.

    Which projects share features with google-antigravity/antigravity-sdk-python?

    Projects with overlapping indexed features include: mervinpraison/praisonai — PraisonAI is an autonomous AI agent platform that coordinates multiple LLM-powered agents for research, planning, and… 1jehuang/jcode — jcode is a framework for developing autonomous AI coding agents that automate software development tasks. It functions… strands-agents/sdk-python — This is an open-source Python SDK for building and orchestrating production-grade AI agents. It provides a unified… nesquena/hermes-webui — Hermes-webui is a self-hosted AI orchestrator and web interface for managing autonomous agents. It serves as a… panaversity/learn-agentic-ai — This project is an educational curriculum and architectural framework for building autonomous AI agents and… omnigent-ai/omnigent — Omnigent is a policy-governed orchestrator for coding agents. It runs several agent harnesses, such as Claude Code and…

    Projects sharing features with Antigravity Sdk Python

    These projects share indexed features with Antigravity Sdk Python. Shared tags can include platform or build tooling; verify the primary use case before treating a result as a replacement.
    • mervinpraison/praisonaiMervinPraison avatar

      MervinPraison/PraisonAI

      9,325View on GitHub↗

      PraisonAI is an autonomous AI agent platform that coordinates multiple LLM-powered agents for research, planning, and execution of complex workflows. It functions as a multi-agent orchestration framework, a workflow builder, and a Model Context Protocol server, while also providing retrieval-augmented generation through vector knowledge bases. Agents can interact via CLI, web, or standardized protocols with sandboxed code execution. The platform distinguishes itself with a rich set of agent communication protocols, including A2A, REST, WebSocket, voice and telephony integration, and MCP, allo

      Pythonagentsaiai-agent-framework
      View on GitHub↗9,325
    • 1jehuang/jcode1jehuang avatar

      1jehuang/jcode

      21,592View on GitHub↗

      jcode is a framework for developing autonomous AI coding agents that automate software development tasks. It functions as an agent orchestrator, tool runtime, and semantic memory engine, enabling the creation of agents that can modify code, run tests, and iterate on their own functionality. The project is distinguished by its use of recursive agent swarming, where a hierarchy of collaborating agents can spawn child agents to decompose complex tasks. It implements a semantic memory system that combines vector-based retrieval with graph-based relationship mapping to maintain context across sess

      Rust
      View on GitHub↗21,592
  • strands-agents/sdk-pythonstrands-agents avatar

    strands-agents/sdk-python

    6,176View on GitHub↗

    This is an open-source Python SDK for building and orchestrating production-grade AI agents. It provides a unified framework for creating conversational agents that can use tools, maintain state, and coordinate across multiple language model providers including OpenAI, Anthropic, Google, Amazon Bedrock, and locally-hosted models. The SDK supports multi-agent orchestration through graphs, teams, and swarms, allowing several specialized agents to collaborate on complex tasks. Agents can be composed as callable tools that other agents invoke, and the framework includes policy handlers that inspe

    Python
    View on GitHub↗6,176
  • nesquena/hermes-webuinesquena avatar

    nesquena/hermes-webui

    14,912View on GitHub↗

    Hermes-webui is a self-hosted AI orchestrator and web interface for managing autonomous agents. It serves as a multi-provider gateway that connects cloud and local large language models, providing a central hub to execute scheduled background jobs, run shell commands, and manage agent memory on private hardware. The system distinguishes itself through a persistent memory manager that utilizes knowledge graphs and markdown files for long-term context across sessions. It features a model context protocol host for extending agent capabilities with standardized tools and supports the orchestratio

    Pythonagentai-agentshermes
    View on GitHub↗14,912
    • Compare all 30 related projects→