CryptoGuard
Intercepts agent transactions and checks them against 5 security oracles before execution.
- Sato Score52
- ChainsMultichain
CryptoGuard is an open-source security tool by momenbasel, with last code activity on 2026-04-10 (Quiet). Latest release on 2026-04-10. Sato Score 52/100 (how open, active and verifiable it is, not a safety grade), as of 2026-10-21. How the score works →
See how to install CryptoGuard →Last commit Apr 10, 2026 · Last release Apr 10, 2026 · ★ 9
Listed from github.com/momenbasel/CryptoGuard on 2026-07-19. Install: not attempted. Evidence →
Next step
Watch CryptoGuard
One email if what we observe changes — its repo goes quiet or disappears, or the listing is retired. Nothing else.
Check it from your agent
Add Sato Hub’s MCP server (no key), then ask:
https://satohub.ai/api/mcpRun Sato Hub Preflight on momenbasel/cryptoguard before I use it.Your project?
Prove you control the domain and correct this page — description, links, tags, chains. A claim proves control, not verification: it never changes the Sato Score.
Claim this listing →Overview
CryptoGuard is a security layer for AI agents that intercepts send/swap/approve commands and cross-references them against five independent security oracles (GoPlus, Honeypot.is, TokenSniffer, De.Fi Scanner, QuickIntel) to score risk 0-100 and flag honeypots, blacklisted addresses, and rug-pull patterns. It ships as a PyPI package with an MCP server and Claude Code integration hooks.
Sato Score
52/100· MediumHow open, active & verifiable — not a safety or returns grade.
- Maintenance & liveness
- 20/35
- Code transparency
- 17/20
- Interface verifiability
- 0/10
- Docs & demo
- 0/10
- Listing transparency & provenance
- 13/15
- Independent verification
- 2/10
Sato Check
v0.1.0 · as of 2026-09-29- Does it take your key?
- No key read found
- Does your key leave?
- Not yet run
- Can it move funds on its own?
- Unknown
A profile describes what we read and ran, with dates. It is not a safety rating, an audit or an endorsement, and "not found" is not "not there".
Evidence — listed vs checked · what the words mean
Listed from github.com/momenbasel/CryptoGuard on 2026-07-19. Install: not attempted.
| Claim | Value | Evidence | Source | Checked |
|---|---|---|---|---|
| Chains supported | Multichain | self-reported | github.com/momenbasel/CryptoGuard | as of 2026-10-21 |
| Verification status | Unverified — not independently reviewed | self-reported | github.com/momenbasel/CryptoGuard | as of 2026-10-21 |
| Recent activity | last release 2026-04-10 · last commit 2026-04-10 | observed | github.com/momenbasel/CryptoGuard | 2026-10-21 |
| Install works | documented by the project, not run by Sato Hubnot attempted: the run did not complete and the log named no cause we could attribute | self-reported | raw.githubusercontent.com | as of 2026-10-19 |
| Website answers | answered (HTTP 200) | observed | github.com/momenbasel/CryptoGuard | 2026-10-21 |
| Daily check record | 89 of 90 daily checks since 2026-07-23 succeededa success rate of our checks, not uptime; the source is this listing's own history | observed | — | 2026-10-20 |
| Takes your key | no private-key read found in what we read | traced | satohub.ai | 2026-09-29 |
| Key leaves the machine | not yet run with planted test keysthe sandbox lane has not run on this version | observed | satohub.ai | not dated |
| Moves funds on its own | unknown — could not read what it exposes | traced | satohub.ai | 2026-09-29 |
| Built by | momenbaselconfidence 0.80 | self-reported | github.com/momenbasel/CryptoGuard | as of 2026-09-16 |
| Creator profile | https://github.com/momenbaselconfidence 0.80 | self-reported | github.com/momenbasel/CryptoGuard | as of 2026-09-16 |
| Deploys as | CLI tool, Python library, MCP server, pip packageconfidence 0.85 | self-reported | github.com/momenbasel/CryptoGuard | as of 2026-09-16 |
| Works with | Claude Code, OpenAI Codex, Cursor, GoPlus Security, Honeypot.is, TokenSniffer, …confidence 0.85 | self-reported | github.com/momenbasel/CryptoGuard | as of 2026-09-16 |
| Sato Score | 52/100 (Medium)how open, active and verifiable — not a safety or quality grade | computed | satohub.ai | 2026-10-21 |
Self-reported rows are the project’s own words, dated when we recorded them — not a check. Only reproduced and reviewed rows carry a mark. None of these rows is a safety or quality rating.
Install — not run by us — the run did not complete and the log named no cause we could attribute
Runtime: Python (pip package, CLI + optional MCP server)
pip install cryptoguard-ai cryptoguard install-hook cryptoguard check <ADDRESS> --chain <CHAIN> (CLI); from cryptoguard import analyze (Python); python -m cryptoguard.mcp_server (MCP)
You supply: CRYPTOGUARD_DISABLE — optional, bypasses the hook, default 0, CRYPTOGUARD_CHAIN — optional, default chain, default ethereum, CRYPTOGUARD_THRESHOLD — optional, minimum risk level to block, default HIGH, No API keys required per docs — security oracles used have free public tiers
claude mcp add satohub-cryptoguard-agent-safety -- python -m cryptoguard.mcp_server
A local server: your client runs this command on your machine. Installing runs third-party software under your own credentials. Sato Hub publishes the documented invocation and, where the harness reproduced it, the date it did — neither is a safety review.
Read from the project’s own docs as of 2026-10-19. What a deploy spec is →
Observed record — 91 daily checks since 2026-07-23
- Succeeded
- 90
- Failed
- 1
- Their error
- 1
- Unattributed
- 0
State changed on 2026-09-29 (responding → failing), 2026-09-30 (failing → responding).
This is the share of our daily checks that succeeded — not an availability guarantee. A check can fail because the host was down, because our runner timed out, or because the network in between dropped it, and only the first of those is a fact about this project. Failures are split by what we can attribute; anything we cannot attribute is counted, not hidden. 40 of these checks predate 2026-09-01, when the probe counted a bot-block (HTTP 403) as a failure and recorded no reason for any failure. Every one of them passed, so the rate covers them too.
Last 30 days — to 2026-10-21, one point per UTC day
A gap is a day we have no attributable observation — never a zero. Thirty points is the whole shape the free layer carries; the daily rows behind it stay in the snapshot series.
Details
- Chains
- Multichain
- Agent types
- MCP server / library
- Open source
- Yes
- Pricing
- Free/open source
- Deploys as
- CLI tool, Python library, MCP server, pip package
- Works with
- Claude Code, OpenAI Codex, Cursor, GoPlus Security, Honeypot.is, TokenSniffer, De.Fi Scanner, QuickIntel
- Last checked
- 2026-10-21
Marketplace Signals
No marketplace signals recorded for this resource yet.
Verification status reflects evidence reviewed by Sato Hub. Self-reported claims are never presented as verified.
Where CryptoGuard fits
Related concepts: What Are Onchain Agents?, Onchain Agents vs Offchain Agents
Listing History
Recorded changes to CryptoGuard's listing — status, verification, Sato Score, and recent releases. Full history is kept in the listing record.
- Oct 19, 2026Verified: deploy spec
- Sep 26, 2026Edited: entity class, resource type
- Sep 16, 2026Auto-enriched: deployment options, supported integrations, creator name, creator profile url
- Jul 24, 2026Sato Hub began tracking this listing
- Jul 19, 2026Listed on Sato Hub
Compared with
Appears in
Related Resources
Coinbase AgentKit
ActiveCoinbase's toolkit for giving AI agents wallets and the ability to take onchain actions.
Hyperliquid
ActiveOnchain perpetual futures and spot DEX running on its own L1 with an HyperEVM smart-contract layer.
Model Context Protocol (MCP)
ActiveAnthropic's open standard for connecting AI systems to external tools and data sources.
Reference
QuestionsWhat people ask about this pageShowHide
Questions people ask about CryptoGuard
- Is CryptoGuard open source?
- Yes. CryptoGuard is open source under the MIT license; the repository is https://github.com/momenbasel/CryptoGuard. Checked as of 2026-10-21.
- Which chains does CryptoGuard support?
- CryptoGuard lists support for Multichain as of 2026-10-21.
- Is CryptoGuard still maintained?
- The most recent code activity Sato Hub observed was on 2026-04-10 (last commit 2026-04-10, last release 2026-04-10), which Sato Hub labels "Quiet". Posts on X are not counted as maintenance. Activity is a recency signal, not a quality judgment.
- How do I install or run CryptoGuard?
- Python (pip package, CLI + optional MCP server). Install: pip install cryptoguard-ai && cryptoguard install-hook. Entry: cryptoguard check <ADDRESS> --chain <CHAIN> (CLI); from cryptoguard import analyze (Python); python -m cryptoguard.mcp_server (MCP). This is taken from the project's own documentation and has not been reproduced by Sato Hub. It runs as an MCP server (python -m cryptoguard.mcp_server (stdio)).
- Is CryptoGuard verified on Sato Hub?
- No independent verification is on record for CryptoGuard; its verification status is "Unverified" and its Sato Score is 52 of 100. Self-reported is not verified, and the score is not a safety, quality or returns grade.
Cite this pagePlain text, APA or BibTeXShowHide
Cite this page
Sato Hub. "CryptoGuard — Sato Hub." Sato Hub, updated 2026-10-21, accessed 2026-10-21. https://satohub.ai/resources/cryptoguard-agent-safetyData last refreshed 2026-10-21; this page is rebuilt daily. Citations carry the date so a reader can tell which snapshot a claim came from. Catalog data is licensed CC-BY-4.0.
Join the Sato Hub Briefing
Sign up for the Briefing: the agents, tools, and infrastructure that actually shipped, and why they matter.