Runtime governance proxy for MCP tool calls. Blocks prompt injection and capability abuse before tool results reach your agent.
arc-gate-mcp sits between your MCP client and any MCP server. Every tool result passes through Arc Gate governance before reaching your agent — blocking prompt injection attacks that exploit the MCP trust boundary.
Related MCP server: MCP Fetch
pip install arc-gate-mcparc-gate-mcp --upstream "uvx mcp-server-fetch" --policy balanced{
"mcpServers": {
"arc-gate": {
"command": "uvx",
"args": ["arc-gate-mcp", "--upstream", "uvx mcp-server-fetch", "--policy", "browser_agent"]
}
}
}balanced — general purpose
browser_agent — web browsing agents
finance_agent — financial data agents
rag_assistant — document retrieval agents
strict — maximum enforcement
AGPL-3.0
arc-gate-mcp is available as a template on Heym — an enterprise agent platform. Try the Governed Web Research Agent template to see Arc Gate MCP in action.
Calls the 'fetch' tool on the upstream MCP server. Pass arguments as keyword arguments matching the upstream tool's input schema. For the fetch tool: url (required, string) is the URL to retrieve; max_length (optional, int) limits response size; raw (optional, bool) returns HTML instead of markdown. Arc Gate inspects all results for prompt injection before they reach your agent. Policy: balanced. No authentication required. Blocked results return an error string.
| Name | Required | Description | Default |
|---|---|---|---|
| kwargs | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It discloses that Arc Gate inspects results for prompt injection, blocked results return errors, and the raw option affects output format. Key behaviors are transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and front-loaded, covering the core action and parameter details in one paragraph. Every sentence adds value, though slightly more structured formatting could improve readability.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the presence of an output schema (not shown), return values are not needed. The description covers usage, parameters, security, and failure modes. Missing details like rate limits or idempotency are acceptable for a fetch tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has a single generic 'kwargs' string with 0% description coverage. The description compensates fully by explaining that kwargs represent keyword arguments and detailing the actual parameters: url (required, string), max_length (optional, int), raw (optional, bool).
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool fetches a URL via an upstream MCP server, listing parameters like url, max_length, and raw. It is specific and distinguishes the action, though no siblings exist to differentiate.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides context like 'No authentication required' and 'Policy: balanced', but lacks explicit guidance on when to use this tool versus alternatives. Since no sibling tools are defined, this is acceptable but could be more directive.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0fetchScored across 1 tool
Only one tool exists, so there is no possibility of ambiguity.
The single tool is named 'fetch', following a clear verb-based naming convention.
One tool is entirely appropriate for a server designed as a secure proxy for the fetch operation.
The server fully covers its stated purpose of fetching URLs with security inspection; no additional tools are needed.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Security & DLP proxy for MCP: tool-poisoning scans, PII redaction on tool args/results. Beta.
Security firewall for AI agents — scans MCP calls for injection, secrets, and risks.
MCP gateway with runtime security policy, tool-call-level control, and audit of agent actions.
MCP enforcement layer that intercepts AI agent actions and blocks rule violations before execution.
