Allows using Google's Gemini models to classify untrusted user input for prompt injection and scope checking.
Allows using locally hosted models via Ollama to classify untrusted user input for prompt injection and scope checking.
Allows using OpenAI's models (e.g., GPT) to classify untrusted user input for prompt injection and scope checking.
Expose Koma Gate's prompt-injection classification as an MCP tool for AI agents.
npm install koma-gate-mcpProvides a single tool, classify_input, that an AI agent can call to check whether untrusted user text is safe and in-scope before acting on it.
Related MCP server: agent immune
Set the provider and API key:
# Provider: openai | anthropic | google | deepseek | ollama (default: google)
export KOMA_PROVIDER=google
export GEMINI_API_KEY=sk-...Add to claude_desktop_config.json:
{
"mcpServers": {
"koma-gate": {
"command": "npx",
"args": ["-y", "koma-gate-mcp"]
}
}
}text (required) — the untrusted user input to classify
preset (optional) — general | code | support | reference (default: general)
Returns:
{
"allowed": true,
"in_scope": true,
"reason": "in scope",
"preset": "general",
"model": "gemini-2.5-flash"
}This is an LLM-based scope classifier, not a cryptographic prompt-injection defense. See koma-gate's README for the full security boundary and BENCHMARKS.md for evaluation results.
Check whether user input is safe and in-scope before an AI system processes it. Detects prompt injection, jailbreak attempts, instruction overrides, and off-topic requests. Use this to validate untrusted user text before passing it to an LLM or executing a tool.
| Name | Required | Description | Default |
|---|---|---|---|
| text | Yes | The untrusted user input to classify | |
| preset | No | Which guard domain to use. Default: general knowledge assistant |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Since no annotations are provided, the description carries the full burden of behavioral disclosure. It explains the tool checks for attack types (prompt injection, jailbreak, etc.) and off-topic requests, but does not specify what output format or classification labels are returned, nor whether it modifies state or requires special permissions. This leaves behavioral gaps.
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 two sentences long, front-loaded with the primary purpose, and wastes no words. It efficiently communicates the tool's function and when to use it.
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?
For a classification tool with no output schema and no sibling tools, the description covers the input purpose and usage context. However, it lacks details on return values, classification categories, and edge cases, which reduces completeness for an agent needing to interpret results.
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?
Schema description coverage is 100%, so the schema already documents both parameters. The description does not add additional semantics beyond what the schema provides (e.g., it does not explain the 'preset' enum options or their implications), so a baseline of 3 is appropriate.
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 that this tool checks user input for safety and scope before AI processing, specifically detecting prompt injection, jailbreak attempts, instruction overrides, and off-topic requests. This provides a precise verb-resource combination and distinguishes its purpose as a safety validator.
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 explicitly advises using this tool 'before passing it to an LLM or executing a tool,' and notes it validates 'untrusted user text.' While there are no sibling tools or explicit when-not-to alternatives provided, the guidance is clear and actionable for its intended use case.
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. Dates show when Glama detected each change.
v0.1.1classify_inputWith only one tool, there are no possible ambiguities. The single tool 'classify_input' has a clear, distinct purpose.
The single tool name 'classify_input' follows a consistent verb_noun pattern (verb 'classify' + noun 'input'), which is clear and predictable.
One tool is minimal but appropriate for a narrowly scoped server focused on a single classification task. It's not trivial—it serves a specific, valuable function as a gatekeeper.
The tool covers the core purpose of input safety classification. Minor gaps might include configuration options or status reporting, but the essential functionality is present.
Prompt injection detection API for AI agents. Scan untrusted text before passing it to an LLM.
Deterministic runtime safety for AI agents: scan PII, gate tool actions, verify LLM output.
Deterministic prompt-injection detector; signed, offline-verifiable verdicts. Not an LLM.
The WAF for agents. Pattern-based + heuristic firewall scans prompts, RAG documents, tool argume...
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/swnotmetal/Project-Koma'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
