A previous release had provenance metadata, but this release does not. View details.
ATR - Agent Tool Registry
A type-safe, decentralized tool registry for autonomous agents. Part of the Agent OS ecosystem.
Why This Exists
Most agent frameworks hardcode tools directly into their runtimes. This creates tight coupling: add a new capability, restart the entire system. Change a function signature, update dozens of agents. Scale by addition leads to fragility.
We built atr because tool registration should not require restarting your infrastructure.
The Agent Tool Registry decouples tool providers from tool consumers. Agents discover capabilities at runtime through a standardized interface. We subtract the dependency between agent logic and tool implementation to add scale.
This is Scale by Subtraction applied to the agent capability layer.
Installation
pip install agent-tool-registry
For sandboxed execution with Docker:
pip install agent-tool-registry[sandbox]
Quick Start
Register a tool in 5 lines:
import atr
@atr.register(name="calculator", tags=["math"])
def add(a: int, b: int) -> int:
"""Add two numbers."""
return a + b
Discover and execute:
tool = atr.get_tool("calculator")
schema = tool.to_openai_function_schema() # OpenAI-compatible
func = atr.get_callable("calculator")
result = func(a=5, b=3) # Returns 8
# Or use sandboxed execution (recommended for untrusted code)
from atr import DockerExecutor
docker_exec = DockerExecutor()
result = atr.execute_tool("calculator", {"a": 5, "b": 3}, executor=docker_exec)
Sandboxed Execution
NEW: ATR now supports sandboxed execution using Docker containers. This is essential for running untrusted or agent-generated code safely.
Why Sandboxed Execution?
SDLC agents and LLMs may generate Python or Bash scripts that you cannot safely run directly on your host machine. Sandboxed execution provides:
- Isolation: Code runs in ephemeral containers, completely isolated from your host
- Security: No network access, memory limits, automatic cleanup
- Safety: Protects against malicious code, resource exhaustion, and unintended side effects
Usage
import atr
from atr import DockerExecutor
# Register a tool
@atr.register(name="processor", tags=["data"])
def process_data(numbers: list) -> int:
"""Process data safely in a sandbox."""
return sum(numbers)
# Option 1: Direct execution (NOT sandboxed - trusted code only)
result = atr.execute_tool("processor", {"numbers": [1, 2, 3, 4]})
# Option 2: Sandboxed execution (RECOMMENDED for untrusted code)
docker_exec = DockerExecutor()
result = atr.execute_tool(
"processor",
{"numbers": [1, 2, 3, 4]},
executor=docker_exec,
timeout=30
)
Execution Modes
| Feature | LocalExecutor | DockerExecutor |
|---|---|---|
| Speed | Fast | Slower |
| Security | No isolation | Full isolation |
| Network | Full access | Disabled |
| Use Case | Trusted code | Untrusted code |
See examples/sandbox_demo.py for complete examples.
Architecture
atr sits in Layer 2 (Infrastructure) of the Agent OS stack.
Responsibility: Tool registration, discovery, and schema generation.
Not responsible for: Tool execution (handled by the Agent Control Plane).
Design
- Registry: In-memory dictionary-based lookup (local or distributed).
- Decorator:
@atr.register()extracts type signatures and validates strict typing. - Spec: Pydantic schema enforcing inputs, outputs, side effects, and metadata.
- Schema Export: Converts to OpenAI, Anthropic, and other LLM function-calling formats.
The registry stores specifications, not callables. Execution happens in the control plane with proper error handling and observability.
The Ecosystem Map
ATR is one component in a modular Agent OS. Each layer solves a specific problem:
Primitives (Layer 1)
- caas - Context-as-a-Service: Manages agent memory and state.
- cmvk - Context Merkle Verification Kit: Cryptographic verification of context integrity.
- emk - Episodic Memory Kit: Long-term memory storage and retrieval.
Infrastructure (Layer 2)
- iatp - Inter-Agent Trust Protocol: Secure message authentication.
- amb - Agent Message Bus: Decoupled event transport.
- atr - Agent Tool Registry: Tool discovery and schema generation (you are here).
Framework (Layer 3)
- agent-control-plane - The Core: Agent orchestration and lifecycle management.
- scak - Self-Correction Agent Kit: Automated error recovery and learning.
Citation
If you use ATR in research, please cite:
@software{atr2024,
title={ATR: Agent Tool Registry},
author={Siddique, Imran},
year={2024},
url={https://github.com/imran-siddique/atr},
note={Part of the Agent OS ecosystem}
}
License
MIT License - See LICENSE for details.
Repository: https://github.com/imran-siddique/atr
Documentation: https://github.com/imran-siddique/atr#readme
Issues: https://github.com/imran-siddique/atr/issues
Metadata
Release files for agent-tool-registry 0.3.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Loss of provenance
These files do not have provenance metadata, despite files in a prior release including this data.
We recommend further investigation before installing these files. View details.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agent_tool_registry-0.3.0.tar.gz | 56.2 kB | Without provenance | Details for agent_tool_registry-0.3.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agent_tool_registry-0.3.0-py3-none-any.whl | Python 3 | none | any | Details for agent_tool_registry-0.3.0-py3-none-any.whl |
Total release size: 105.8 kB
Release files / agent_tool_registry-0.3.0.tar.gz
Loss of provenance
This file belongs to a release without provenance metadata, despite a prior release including this data.
We recommend further investigation before installing this file. View details.
| Download URL | agent_tool_registry-0.3.0.tar.gz |
|---|---|
| Size | 56.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
3a04bf5195e8dc02c8abc4084fa55f909c60acf53e81552bd3e680c2b8e83a55
|
|
BLAKE2b-256 checksum How to use checksums |
6cc81185bf94cf60cb444366033527cd828cf7d97c0050dea6775872c6b8a409
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.9
|
Release files / agent_tool_registry-0.3.0-py3-none-any.whl
Loss of provenance
This file belongs to a release without provenance metadata, despite a prior release including this data.
We recommend further investigation before installing this file. View details.
| Download URL | agent_tool_registry-0.3.0-py3-none-any.whl |
|---|---|
| Size | 49.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
7f33c6ed464d6014ac8a4db732d34f7e47f9a21ee34ac1e587ff27dc39e7c133
|
|
BLAKE2b-256 checksum How to use checksums |
81382d6cda24227271c982cd88d67076605b84485745390358ba74761d7cc3e2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.9
|
Security for agent-tool-registry 0.3.0
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.
Loss of provenance
A previous release of these files had provenance metadata, but this release does not.
While this could reflect a routine change by the project maintainers, it may also indicate that the release files have been compromised.
Release comparison
- Removed 0.1.0 2 files With provenance
- Added 0.3.0 2 files Without provenance
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.