Skip to main content

UnexHub Agent Runtime

Build a chat Agent by supplying instructions, knowledge, and optional Python business logic. The official Runtime handles the unex-agent-v1 container API, a default model/tool loop, streaming events, user questions, and cancellation. UnexHub supplies model and MCP capabilities for each authorized Run; UNH executes the local tools.

Requires Python 3.11–3.14. The package includes the CLI, starter template, dependency lock, and three examples. Developers do not need to clone the Runtime repository or build it from source.

Create an Agent

Once the selected version is available on PyPI:

pip install unexhub-agent
unexhub-agent init my-agent

Or let uv manage the tool environment:

uvx --from unexhub-agent==0.1.1 unexhub-agent init my-agent

Initialization makes no network requests. requirements.runtime.txt pins the CLI's exact Runtime version; image builds resolve that version from the configured package index. Upgrading your local CLI does not alter existing projects or released images.

Before publication, or to carry the Runtime with the project, install a maintainer-provided wheel and initialize with it:

python -m pip install /path/to/unexhub_agent-0.1.1-py3-none-any.whl
unexhub-agent init my-agent \
  --runtime-wheel /path/to/unexhub_agent-0.1.1-py3-none-any.whl

The wheel must match the installed CLI version. It is copied into vendor/ and referenced with a relative path. Fully disconnected builds also require a source or cache for third-party and business dependencies.

Add your business rules

Edit agent.yaml, instructions.md, and knowledge/. The default entrypoint: null uses the official loop. To add Python logic, set entrypoint: agent:run and implement:

from unexhub_agent import AgentContext


async def run(ctx: AgentContext) -> str:
    await ctx.emit("Starting the task", kind="run.progress")
    return await ctx.run_agent()

Use --example browser-assistant, --example file-organizer, or --example monitoring to initialize a complete example. The local wheel option can be combined with any example. Browser tools must be active in UNH; artifact publishing requires the platform artifact service.

Build and cloud-test

cd my-agent
unexhub-agent check-config --config agent.yaml
docker build --platform linux/amd64 -t my-agent:0.1.1 .

The image listens on 0.0.0.0:8080 and provides /health. Use the existing UnexHub Mode A/B image validation, deployment, self-test, and review workflow. Configure chat or hybrid interaction, protocol unex-agent-v1, and the Runtime version inside the image. Connect an online UNH device signed into the same account and environment, then send the task in UNH.

The platform owns container creation, renewal, stopping, and cleanup. A test Run borrows an existing cloud self-test environment and does not destroy or extend it when the task ends. This package does not connect UNH directly to an Agent on the developer's machine. Container files are temporary; persist business data explicitly. Unknown outcomes must be reconciled rather than blindly replayed.

UnexHub · Developer Center

Container egress proxies

Model, MCP, and artifact HTTP clients use HTTPX trust_env=True. They honor the platform-injected HTTP_PROXY, HTTPS_PROXY, ALL_PROXY, and NO_PROXY settings. Keep loopback addresses in NO_PROXY for local health checks and fixtures. The container network policy must also allow the actual UnexHub Broker hostname. Rebuild existing Agent images with the updated pinned Runtime or a maintainer-provided wheel to apply this change.

License

The Runtime, starter template, and examples are licensed under the Apache License, Version 2.0. Generated projects retain Runtime license and attribution files in vendor/. Independent business code and third-party dependencies retain their respective licenses.

Metadata

Release files for unexhub-agent 0.1.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for unexhub-agent 0.1.1
File Size Uploaded Provenance Actions
unexhub_agent-0.1.1.tar.gz 175.0 kB Without provenance Details for unexhub_agent-0.1.1.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for unexhub-agent 0.1.1
File Interpreter ABI Platform Actions
unexhub_agent-0.1.1-py3-none-any.whl Python 3 none any Details for unexhub_agent-0.1.1-py3-none-any.whl

Total release size: 231.2 kB

Release files / unexhub_agent-0.1.1.tar.gz

Download URL unexhub_agent-0.1.1.tar.gz
Size 175.0 kB
Tags Source
SHA-256 checksum
How to use checksums
6de0ffc2cad5b89945d1b95a54db80d2ff48efebd834bbe3e18afffebb88210b
BLAKE2b-256 checksum
How to use checksums
4cfea18ba96e8d9db3c06d319ff4166749ac611af6e7ae9342c8aa474e04c9c3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.10.6 {"installer":{"name":"uv","version":"0.10.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release files / unexhub_agent-0.1.1-py3-none-any.whl

Download URL unexhub_agent-0.1.1-py3-none-any.whl
Size 56.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
a0e5c6152880f1ac59b47582dcc2c1e19958087e4917e5d46eec79b46bef2839
BLAKE2b-256 checksum
How to use checksums
63953c01bddde0c62ea65509a3e451c044ba0f5f5a32015d98379bd5f0e50df6
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.10.6 {"installer":{"name":"uv","version":"0.10.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Security for unexhub-agent 0.1.1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page