🤺 Fence
The Bloat Moat! A lightweight, production-ready library for LLM communication and agentic workflows. Born from the need for something simpler than LangChain, Fence gives you powerful LLM orchestration without the heavyweight dependencies.
Think of it as the Swiss Army knife for LLM interactions—sharp, reliable, and it won't weigh down your backpack (or your Docker image).
🤔 Why Fence?
The short answer: By accident.
The slightly longer answer: LangChain used to be (is?) a pretty big package with a ton of dependencies. Great for PoCs, but in production? Not so much.
The problems we faced:
- 🐘 It's BIG. Takes up serious space (problematic in Lambda, containers, edge environments)
- 🌀 It's COMPLEX. Overwhelming for new users, hard to debug in production
- 💥 It BREAKS. Frequent breaking changes, version jumps that made us cry
As a result, many developers (especially those in large production environments) started building lightweight, custom solutions that favor stability and robustness over feature bloat.
Enter Fence 🤺
We started building basic components from scratch for our Bedrock-heavy production environment. First came the Link class (wink wink), then templates, then agents... and before we knew it, we had a miniature package that was actually fun to use.
Fence strikes the perfect balance between convenience and flexibility.
Note: Fence isn't trying to replace LangChain for complex PoCs. But if you want a simple, lightweight, production-ready package that's easy to understand and extend, you're in the right place.
📦 Installation
pip install fence-llm
That's it. Seriously. No 500MB of transitive dependencies.
🚀 Quick Start
Hello World (The Obligatory Example)
from fence.links import Link
from fence.templates.string import StringTemplate
from fence.models.openai import GPT4omini
# Create a link
link = Link(
model=GPT4omini(),
template=StringTemplate("Write a haiku about {topic}"),
name='haiku_generator'
)
# Run it
output = link.run(topic='fencing')['state']
print(output)
Output:
[2024-10-04 17:45:15] [ℹ️ INFO] [links.run:203] Executing <haiku_generator> Link
Blades flash in the light,
En garde, the dance begins now,
Touch—victory's mine.
Much wow. Very poetry. 🎭
💪 What Can Fence Do?
Fence is built around a few core concepts that work together beautifully:
🤖 Multi-Provider LLM Support
Uniform interface across AWS Bedrock (Claude, Nova), OpenAI (GPT-4o), Anthropic, Google Gemini, Ollama, and Mistral. Switch models with a single line change.
🔗 Links & Chains
Composable building blocks that combine models, templates, and parsers. Chain them together for complex workflows.
👉 Learn about Links & Chains →
🤖 Agentic Workflows ⭐
The crown jewel! Production-ready agents using the ReAct pattern:
Agent- Classic ReAct with tool use and multi-level delegationBedrockAgent- Native Bedrock tool calling with streamingChatAgent- Conversational agents for multi-agent systems
🔌 MCP Integration
First-class support for the Model Context Protocol. Connect to MCP servers and automatically expose their tools to your agents.
🎭 Multi-Agent Systems
Build collaborative agent systems with RoundTable where multiple agents discuss and solve problems together.
🧠 Memory Systems
Persistent and ephemeral memory backends (DynamoDB, SQLite, in-memory) for stateful conversations.
🛠️ Tools & Utilities
Custom tool creation, built-in tools, retry logic, parallelization, output parsers, logging callbacks, and benchmarking.
📚 Documentation
- Models - All supported LLM providers and how to use them
- Links & Chains - Building blocks for LLM workflows
- Agents - ReAct agents, tool use, and delegation
- MCP Integration - Model Context Protocol support
- Multi-Agent Systems - RoundTable and collaborative agents
- Memory - Persistent and ephemeral memory backends
- Tools & Utilities - Custom tools, parsers, and helpers
🎯 Examples
Simple Agent with Tools
from fence.agents import Agent
from fence.models.openai import GPT4omini
from fence.tools.math import CalculatorTool
agent = Agent(
identifier="math_wizard",
model=GPT4omini(source="demo"),
tools=[CalculatorTool()],
)
result = agent.run("What is 1337 * 42 + 999?")
print(result) # Agent thinks, uses calculator, and answers!
BedrockAgent with MCP
from fence.agents.bedrock import BedrockAgent
from fence.mcp.client import MCPClient
from fence.models.bedrock import Claude37Sonnet
# Connect to MCP server
mcp_client = MCPClient(
transport_type="streamable_http",
url="https://your-mcp-server.com/mcp"
)
# Create agent with MCP tools
agent = BedrockAgent(
identifier="mcp_agent",
model=Claude37Sonnet(region="us-east-1"),
mcp_clients=[mcp_client], # Tools auto-registered!
)
result = agent.run("Search for customer data")
Multi-Agent Collaboration
from fence.troupe import RoundTable
from fence.agents import ChatAgent
from fence.models.openai import GPT4omini
# Create specialized agents
detective = ChatAgent(
identifier="Detective",
model=GPT4omini(source="roundtable"),
profile="You are a sharp detective."
)
scientist = ChatAgent(
identifier="Scientist",
model=GPT4omini(source="roundtable"),
profile="You are a forensic scientist."
)
# Let them collaborate
round_table = RoundTable(agents=[detective, scientist])
transcript = round_table.run(
prompt="A painting was stolen. Let's investigate!",
max_rounds=3
)
More examples:
- 📓 Jupyter Notebooks - Interactive tutorials
- 🎬 Demo Scripts - Runnable examples
🤝 Contributing
We welcome contributions! Whether it's:
- 🐛 Bug fixes
- ✨ New features (especially new model providers!)
- 📝 Documentation improvements
- 🧪 More tests
- 🎨 Better examples
Check out CONTRIBUTING.md for guidelines.
📄 License
MIT License - see LICENSE.txt for details.
🙏 Acknowledgments
Inspired by LangChain, built for production, made with ❤️ by developers who got tired of dependency hell.
Now go build something awesome! 🚀
1 maintainer
Metadata
1 maintainer
Release files for fence-llm 2.3.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| fence_llm-2.3.3.tar.gz | 160.7 kB | With provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| fence_llm-2.3.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 298.1 kB
Release files / fence_llm-2.3.3.tar.gz
| Download URL | fence_llm-2.3.3.tar.gz |
|---|---|
| Size | 160.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
14deecd869a1260f5bb4022e0ff6cd7c541a35997b918bf08cb3639059159bd5
|
|
BLAKE2b-256 checksum How to use checksums |
8e8094e71daaeea43f7d6e601702ab9c2bb79a9b1c1753e5def4b99fa016d81d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
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.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Mar 16, 2026.
Transparency logRelease files / fence_llm-2.3.3-py3-none-any.whl
| Download URL | fence_llm-2.3.3-py3-none-any.whl |
|---|---|
| Size | 137.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
288105cf6c18f677b8ea7622c917607c5d89513eff6ed704ebc7df51e1fe8acf
|
|
BLAKE2b-256 checksum How to use checksums |
cd7fba5cf699fe78ca585ce7b5750d14713897251ef2ca20cf4b3613e9c76380
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
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.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Mar 16, 2026.
Transparency logSecurity for fence-llm 2.3.3
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.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for fence_llm-2.3.3.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Mar 16, 2026.
Transparency logA 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.
