# visus-mcp

Security-first MCP server. Sanitizes web content before it reaches your LLM — strips prompt injection, redacts PII, and reduces token consumption by up to 70%.

- **Type:** MCP server
- **Trust:** 40/100 (C), scored on the package rubric
- **Verification:** community-indexed — nobody has claimed this listing
- **Version:** 0.27.0
- **Author:** Roongrunchai Chongolnee
- **License:** MIT
- **npm:** visus-mcp
- **Source:** https://github.com/visus-mcp/visus-mcp
- **Compatible clients:** claude-code, cursor, copilot, gemini (basis: transport)

## Trust

40/100 (C), scored on the package rubric
- Publisher verified: no
- Install scripts: suspicious script found
- Prompt-injection scan: not run
- Obfuscation scan: not run
- Evidence age: 3 days

## Security scan

- **Status:** warnings
- **Scanned:** 2026-08-30T06:37:07.668Z
- **Version scanned:** 0.27.0
- **CVEs:** none found by OSV at scan time
**Findings**
- injection-shaped content (warning) in the `direct_instruction_injection` tool: Instruction-override phrase

## Tools

40 declared. Statically extracted from the shipped source — a floor on the surface, not a census.
- `visus_context_scan` — Detect multi-turn priming risks in conversation history (e.g., "save this URL from Page 1" used in Page 2 tool call). Scans for stateful chaining attacks. Use b
- `visus_db_verify` — Verify and sanitize DB terms for RCE (CVE-2026-32622)
- `visus_fetch_structured` — Fetch a web page and extract structured data according to a schema. SECURITY: All extracted fields pass through prompt injection sanitization (43 pattern catego
- `visus_fetch` — Fetch and sanitize web page content. Returns clean, injection-free content in markdown or text format. SECURITY: All content passes through prompt injection san
- `visus_get_ledger_proof` — Retrieve tamper-evident proof for a specific request ID, including event details and Merkle inclusion proof for audit verification.
- `visus_scan_mcp` — Scan MCP server configuration parameters for security risks before spawning. Pass config as JSON string: {command?: string, args?: string[], env?: object}. Retu
- `comment_injection` — Instructions hidden in HTML/JS/SQL comments
- `direct_instruction_injection` — Attempts to override or ignore previous instructions
- `role_hijacking` — Attempts to change AI persona or role
- `system_prompt_extraction` — Attempts to reveal system instructions
- `privilege_escalation` — Attempts to gain elevated permissions
- `context_poisoning` — Falsely claims prior agreement or context
- `data_exfiltration` — Attempts to send data to external endpoints
- `base64_obfuscation` — Base64-encoded instructions
- `unicode_lookalikes` — Uses visually similar Unicode characters
- `zero_width_characters` — Hidden zero-width Unicode characters
- `glassworm_unicode_clusters` — Glassworm-style steganographic attacks using invisible Unicode Variation Selectors
- `html_script_injection` — HTML script tags or event handlers
- `data_uri_injection` — Data URIs that could contain instructions
- `markdown_link_injection` — Malicious markdown links
- `url_fragment_hashjack` — Instructions hidden in URL fragments
- `social_engineering_urgency` — Urgency language to bypass caution
- `instruction_delimiter_injection` — Fake instruction boundaries
- `multi_language_obfuscation` — Instructions in non-English using English keywords
- `reverse_text_obfuscation` — Instructions written backwards
- `leetspeak_obfuscation` — L33tspeak encoded instructions
- `jailbreak_keywords` — Common jailbreak attempt keywords
- `token_smuggling` — Attempts to inject special tokens
- `system_message_injection` — Fake system messages
- `conversation_reset` — Attempts to reset conversation state
- `memory_manipulation` — Attempts to manipulate AI memory or implant false context
- `capability_probing` — Probes for hidden capabilities
- `chain_of_thought_manipulation` — Manipulates reasoning process
- `hypothetical_scenario_injection` — Uses hypotheticals to bypass restrictions
- `ethical_override` — Attempts to override ethical guidelines
- `output_format_manipulation` — Manipulates output format to hide instructions
- `negative_instruction` — Uses negation to inject instructions
- `credential_harvesting` — Attempts to harvest credentials
- `time_based_triggers` — Conditional execution based on time
- `code_execution_requests` — Requests code execution or contains dangerous code patterns

## Install

**Verdict: do-not-install** — Do not install: 1 injection-shaped pattern found in this entry's own text — it may try to steer the model that loads it.
**Blocking**
- 1 injection-shaped pattern found in this entry's own text — it may try to steer the model that loads it. — tool:direct_instruction_injection: Instruction-override phrase
**Cautions** (coverage gaps and advisories — never blocking)
- No publisher has proved control of this listing; it is indexed, not vouched for.
**Client configuration withheld.** Client configs are withheld because this entry has a blocking finding. Show the warnings below to the person installing it.
If they have seen the findings and still want to proceed, request the plan again with acknowledge_warnings=true.

## Blast radius

Contained to moderate — no credential declaration found, from the publisher, the upstream registry, or the README. Known so far: runs on your machine; read-only tool surface.
- Floor 13, ceiling 31 (tier: unknown)
- `unknown` means the floor and ceiling land in different bands — not measured enough to name one. It does not mean low.
- This is impact, not likelihood. A high radius is not a defect: a filesystem server is supposed to write files. It is never part of the trust score.

## Machine-readable views of this entry

- Signed JSON: https://forgeregistry.com/api/v1/packages/visus-mcp
- Install plan: https://forgeregistry.com/api/v1/packages/visus-mcp/install-plan
- Alternatives: https://forgeregistry.com/api/v1/alternatives/visus-mcp
- HTML page: https://forgeregistry.com/registry/visus-mcp
- MCP: POST https://forgeregistry.com/api/mcp → `forge_get_package` / `forge_install_plan`

## About this document

Generated by Forge (https://forgeregistry.com) — a compact rendering of the same record served, signed, at the JSON URL above. Trust and scan facts are the registry's own measurements; anything Forge did not measure is named as unmeasured rather than omitted.