Simpleflo Utility
Conduit
Local-first context and connectors for grounded AI workflows.
Conduit is a CLI utility for local-first context, managed connectors, and grounded AI workflows. It keeps your data on your terms while making it easier to wire reliable context into the tools you already use.
The problem
AI tools become far more useful when they understand the context you work in. In practice, that context is scattered across documents, notes, folders, and internal references. The usual workaround is to keep uploading files and pasting snippets, which becomes messy and unreliable as the pile grows.
At the same time, connecting AI tools to real systems is still too manual. You install a connector, chase dependencies, edit configs, and repeat the process for each client. This is the kind of undifferentiated setup work that drains momentum.
What Conduit changes
Conduit is designed to make your AI tools feel more personal and more capable without increasing your risk. It helps you manage MCP servers as a guided, secure experience, and it supports a private knowledge base that turns your documents into usable personal context.
Instead of sending every file to every AI tool, Conduit builds a knowledge layer from what you choose to add. That layer is designed to return precise, minimal context so your AI tool gets what it needs without being flooded by irrelevant material.
Conduit is a CLI-first private knowledge base that runs locally and exposes your context to AI tools via an MCP server. It helps AI tools pull just the right context on demand—reducing oversharing and avoiding context bloat. Today, Conduit ships the local KB + MCP connector workflow. What’s actively evolving is the “delight layer”: a more guided setup experience, an experimental macOS GUI companion built on top of the CLI, and future work toward automated MCP connector management across AI clients.
What you can do
- Add files and folders to build a private knowledge base that stays under your control.
- Bring personal context into AI tools through a first-party MCP server that Conduit runs and maintains.
- Install and configure MCP servers with guidance, safety checks, and sensible defaults.
- Configure once and reuse across multiple clients without repeating setup steps.
Safety by default
Conduit treats third-party servers as untrusted by default. It runs them in isolated containers, stores secrets locally in your OS keychain, and requires explicit permission before granting access to sensitive resources. This approach is designed to reduce the risk of running connector code you did not write yourself.
FAQ
Is Conduit a cloud service?
The core experience is designed around local execution and local secret storage. Some clients require a reachable HTTPS endpoint, and Conduit is designed to support that case with clear controls.
Why not set up MCP servers manually?
You can, but you end up repeating the same work across clients and taking on more risk. Conduit is designed to reduce that setup burden while keeping safety as the baseline.
What does “personal context” mean here?
It means a private knowledge base built from your own materials, shaped into usable context that AI tools can query without being overwhelmed by raw files.
Designed for calm execution
Conduit keeps the moving pieces visible: connectors, storage, and retrieval paths. You can verify what data is used, audit the workflow, and keep operations local-first.
Local-first context
Keep the source-of-truth on your machine and control when data syncs out.
Managed connectors
Connect to tools, files, and services with explicit boundaries and auditing.
Grounded workflows
Keep retrieval visible with logs, metadata, and structured paths for review.
CLI-first
Scriptable commands with a deterministic output model for repeatable runs.
Start with the essentials
The docs cover install steps, quickstart flows, CLI commands, and the architecture behind Conduit.