MCP server for Hejdar — runtime policy enforcement for AI agents.
This server exposes hejdar_evaluate as an MCP tool. Any MCP-compatible agent (Claude, ChatGPT, Cursor, custom) can call it to check whether an action is permitted by organizational policy before executing it.
The MCP server is a thin wrapper around the Hejdar API (POST /v1/evaluate). It contains no policy logic — all decisions come from your Hejdar organization's configured policies.
pip install hejdar-mcpOr run directly with uvx:
uvx hejdar-mcpSign up at app.hejdar.com and create an API key in Settings → API Keys.
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"hejdar": {
"command": "uvx",
"args": ["hejdar-mcp"],
"env": {
"HEJDAR_API_KEY": "hejdar_sk_your_key_here"
}
}
}
}Add to your Claude Code MCP settings:
{
"mcpServers": {
"hejdar": {
"command": "uvx",
"args": ["hejdar-mcp"],
"env": {
"HEJDAR_API_KEY": "hejdar_sk_your_key_here"
}
}
}
}export HEJDAR_API_KEY=hejdar_sk_your_key_here
hejdar-mcpRelated MCP server: PolicyGuard
Install: pip install hejdar-mcp or uvx hejdar-mcp
Get an API key — contact us at hello@hejdar.com or visit hejdar.com
Configure your MCP client (see configuration example above)
hejdar_evaluateEvaluate an agent action against your organization's security policies.
Input:
Parameter | Type | Required | Description |
| string | Yes |
|
| string | Yes | Target resource, e.g. |
| string | No | Name of the calling agent, e.g. |
| object | No | Free-form metadata (department, user_id, reason, etc.) |
Output:
{
"decision": "DENY",
"policy_id": "pol_abc123",
"reason": "Deletion of customer data requires manager approval",
"risk_level": "HIGH"
}decision is one of: ALLOW, DENY, WOULD_DENY.
For best results, add this to your agent's system prompt:
You have access to the hejdar_evaluate tool. Before performing any action
that reads, writes, deletes, transfers data, or executes commands on
external systems, you MUST call hejdar_evaluate first.
If hejdar_evaluate returns DENY or WOULD_DENY, do NOT proceed with the
action. Instead, inform the user that the action was blocked by policy
and include the reason provided.Variable | Required | Default | Description |
| Yes | — | Your Hejdar API key |
| No |
| API base URL (for self-hosted) |
API key is read from environment variables only — never hardcoded or exposed in tool I/O
All inputs are validated and sanitized before forwarding to the API
Error responses never leak internal details, API keys, or stack traces
All API calls enforce TLS
git clone https://github.com/ARKALDA/hejdar-mcp.git
cd hejdar-mcp
pip install -e ".[dev]"
pytestMIT
Evaluate an AI agent action against Hejdar security policies BEFORE executing it. Returns ALLOW, DENY, or WOULD_DENY. Call this before any sensitive action (read, write, delete, transfer, execute) to check if the action is permitted by organizational policy. If the decision is DENY, do NOT execute the action.
| Name | Required | Description | Default |
|---|---|---|---|
| action_type | Yes | The type of action the agent intends to perform | |
| resource | Yes | The target resource or system the action applies to, e.g. 'customer_database', 'employee_records', 'email_system' | |
| agent_name | No | Name identifying this agent, e.g. 'hr-assistant', 'finance-bot' | |
| context | No | Optional metadata about the action — department, user_id, reason, data_classification, etc. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It effectively describes the tool's behavior: it performs a pre-execution security evaluation, returns one of three policy decisions, and has a critical safety implication (preventing execution on DENY). It doesn't mention rate limits, authentication needs, or error handling, but covers the core operational behavior well.
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 perfectly structured and concise. The first sentence establishes the core purpose and output. The second sentence provides critical usage guidelines. The third sentence delivers an essential safety instruction. Every sentence earns its place with no wasted words, and the most important information (what it does and when to use it) is front-loaded.
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 security evaluation tool with no annotations and no output schema, the description provides excellent context about its purpose, usage, and behavioral implications. It doesn't describe the return format details (what ALLOW/DENY/WOULD_DENY responses contain) or potential error cases, but covers the essential operational context sufficiently given the tool's critical safety role.
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 schema has 100% description coverage, so the baseline is 3. The tool description doesn't add any parameter-specific information beyond what's already documented in the schema (action_type, resource, agent_name, context). It mentions these parameters implicitly through examples ('read, write, delete, transfer, execute') but provides no additional semantic context.
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's purpose with specific verbs ('evaluate an AI agent action against Hejdar security policies') and resources ('security policies'), and explicitly distinguishes its role as a pre-execution check. It identifies the exact function (policy evaluation) and output (ALLOW, DENY, WOULD_DENY), leaving no ambiguity about what this tool does.
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 explicit guidance on when to use this tool ('before any sensitive action') and what to do based on the outcome ('if the decision is DENY, do NOT execute the action'). It lists specific action types (read, write, delete, transfer, execute) that should trigger its use, offering clear operational instructions despite no sibling tools for comparison.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
With only one tool, there is no possibility of ambiguity or overlap with other tools. The tool's purpose is clearly defined as evaluating AI agent actions against security policies, making it distinct and unambiguous in isolation.
A single tool inherently has perfect naming consistency since there are no other tools to compare against. The name 'hejdar_evaluate' follows a clear pattern of server prefix and action, which would be consistent if more tools existed.
A single tool is too few for a server that claims to handle security policy evaluation across various actions (read, write, delete, transfer, execute). This minimal set forces agents to rely solely on this one tool without dedicated tools for different policy aspects or actions, making the scope feel incomplete and thin.
The server's domain appears to be security policy evaluation for AI actions, but with only one tool, there are significant gaps. It lacks tools for managing policies, querying specific rules, or handling different types of security checks, which limits agents to a single evaluation call without supporting operations for a comprehensive workflow.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Deterministic runtime safety for AI agents: scan PII, gate tool actions, verify LLM output.
Security gateway for AI agents: policy, approval, and audited execution, no secrets shared.
See, price, and control every tool call your AI agents make: policy checks, cost, and audit tools.
Zero-trust gateway for AI agents: score tool calls, verify agent cards, enforce policy, audit.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ARKALDA/hejdar-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
