Agentkit Platform Python SDK and Starter Toolkit

Launch your local agent on Volcengine AgentKit Platform as a fully managed service.

Volcengine AgentKit ◆ Documentation ◆ Samples ◆ PyPI Package ◆ GitHub Repository

AgentKit is a developer platform by Volcengine that supports the building, deployment, and operation of AI Agents. It lowers the entry barrier for developers and enterprises by providing essential infrastructure beyond the model—including security, built-in tools, memory, knowledge, monitoring, and evaluation. This empowers enterprises to efficiently build, deploy, and operate complex, intelligent, enterprise-grade Agents. The platform also includes a Python SDK and a Starter Toolkit to help developers build, deploy, publish, and manage Agent applications through an SDK and CLI.

AgentKit includes the following modular Services that you can use together or independently:

AgentKit Runtime is a fully managed service that provides a secure, isolated environment for running AI Agents. It supports the deployment of Agents built with any framework and language, and provides a set of APIs for interacting with the Agents.

Runtime Quick Start

AgentKit Tools is a service that provides a set of built-in tools for AI Agents. It supports the execution of common tasks, such as data retrieval, web search, and code execution, in a secure and scalable manner.

Tools Quick Start

AgentKit Memory is a service that provides a persistent storage solution for AI Agents. It supports the storage of Agent states, memories, and other data in a secure and scalable manner.

Memory Quick Start

AgentKit Knowledge is a service that provides a knowledge base solution for AI Agents. It supports the storage of Agent knowledge, facts, and other data in a secure and scalable manner.

Knowledge Quick Start

AgentKit MCP is a service that provides a set of tools for managing AI Agents. It supports the deployment, configuration, and monitoring of Agents in a secure and scalable manner.

MCP Quick Start

Install the latest stable version:

pip install agentkit-sdk-python

For testing new features or bug fixes before they're officially released:

# Install the latest pre-release version
pip install --pre agentkit-sdk-python

# Or install a specific development version
pip install agentkit-sdk-python==1.0.0.dev1

Note: Development versions may contain bugs and are not recommended for production use.

Stable releases are published from GitHub tags. The source tree keeps the release version in pyproject.toml and agentkit/version.py, and the GitHub workflow verifies that the pushed tag matches those files before building and publishing the package.

The release workflow must already exist on main before you push the release tag. For 0.7.5, merge the tag-release automation change first, then create and push v0.7.5.

Use the following workflow for a stable release:

git checkout main
git pull github main
git tag v0.7.5
git push github v0.7.5

When the tag reaches GitHub, .github/workflows/release-pypi.yml will:

• verify the tag is reachable from main
• verify that pyproject.toml and agentkit/version.py match the tag version
• build the wheel and source distribution from the tagged commit
• verify the built artifacts contain the expected version metadata
• publish the release to PyPI using the GitHub Actions secret PYPI_API_TOKEN

The internal local_build.py helper is a separate local-only workflow and is not used by the GitHub tag release.

Before using the GitHub tag release workflow, configure a project-scoped PyPI API token in the repository's GitHub Actions secrets as PYPI_API_TOKEN.

This project takes security seriously. For vulnerability reporting and supported versions, see SECURITY.md.

This project is licensed under the Apache 2.0 License.