edwang2006

edwang2006 / fides_protocol

Public

Fides — The Trust Layer for AI Agents Fides sits between you and your AI agent. Every tool call is intercepted, verified against your intent constraints, and cryptographically logged on Solana. You see a dashboard. Your agent sees nothing different. The blockchain sees everything.

10
0
100% credibility
Found May 06, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

Fides is an open protocol that adds a trust layer to AI agents by logging actions, enforcing user-defined rules before risky moves, and generating verifiable proofs of compliant behavior.

How It Works

1
🔍 Hear about Fides

You learn about Fides, a safety shield that watches your AI helper to make sure it only does what you allow.

2
💻 Set up on your computer

You easily start the protector program and open a simple control panel on your screen.

3
⚙️ Set your rules

You pick simple no-go rules like 'no file writing' or 'no secret words' to keep things safe.

4
🤖 Connect your AI helper

Your AI assistant now runs through Fides, so every action gets checked first.

5
📊 Watch it protect you

The dashboard shows live what happened, any blocks, and keeps everything clear at a glance.

6
🔒 Review the record

You check the neat log of actions to see proof that your rules were followed.

✅ Use AI worry-free

Now you enjoy your AI's help with full confidence, knowing it's always under your control.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is fides_protocol?

Fides Protocol is a Rust-built trust layer that sits between you and AI agents, intercepting every tool call to check user-defined constraints—like blacklisting write_file tools or keyword blocks in args—before execution, then cryptographically logging summaries on Solana. Users get a React dashboard showing operations, blocks, and stats from local JSONL logs, while agents see nothing different via an MCP proxy. It tackles rogue agent behavior in workflows touching files, APIs, or decisions, providing auditable proof without exposing private context.

Why is it gaining traction?

Zero-disruption proxy mirrors upstream MCP servers, enforcing constraints at runtime with instant blocks or passes. ZKP circuits prove scope and frequency compliance on-chain, blending Web2 dashboard UX with blockchain tamper-evidence. Developers hook it via quick npm commands, gaining verifiable logs against intent violations without agent rewrites.

Who should use this?

AI builders integrating MCP tools for filesystem or API access, needing runtime guards against overreach. Teams at fides trustees mauritius or fides trustees red bull prototyping agent automation on sensitive data. Local agent runners wanting a dashboard view of calls, blocks, and Solana-anchored proofs.

Verdict

Promising MVP for agent trust, but 10 stars and 1.0% credibility signal early days—strong README quickstart and bilingual docs help, though sparse tests and manual ZKP setup limit polish. Prototype for MCP agents now; wait for auto-proofs and mainnet before betting big.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.