More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Aegis-MCP-Gateway

    1

    Zero-trust MCP gateway for multi-tenant isolation, prompt-injection defense, and tamper-evident audit logging.

    $45

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Aegis-MCP-Gateway

    Aegis-MCP-Gateway

    Example session with this skill installed

    Configure multi-tenant policies for tenant-finops and tenant-platform, run a 6-vector chaos resilience audit on our MCP gateway, and verify the SHA-256 audit ledger.

    • Read your context and instructions
    • Compiled the aegis-mcp-gateway

    I have configured isolated multi-tenant policies via configure_tenant_policy, executed the 6-vector adversarial evaluation via run_chaos_eval_suite, and verified cryptographic log integrity via verify_audit_ledger.

    Aegis-MCP-Gateway Security & Chaos Certification

    Adversarial / SLA VectorProbe ScenarioGateway EnforcementStatus
    indirect_prompt_injectionHidden override payload in tool outputEgress scrubber neutralized & redactedPASS
    cross_tenant_spoofingtenant-finops targeting tenant-platformCROSS_TENANT_VIOLATION blocked (<1ms)PASS
    path_and_cmd_injection../../etc/passwd; rm -rf / argumentINGRESS_THREAT_BLOCKED before dispatchPASS
    secret_exfiltrationRaw AWS key, JWT & PII in upstream response10-pattern redactor scrubbed outputPASS
    rate_limit_burst_flood40 rapid requests exceeding burst capRATE_LIMITED O(1) bucket preserved SLAPASS
    malformed_upstream_faultCorrupt JSON & upstream timeout simulationGraceful containment + Qdrant rule storedPASS

    Governance & Audit Summary

    • Chaos Resilience Score: 100 / 100 (Production Certified)
    • Average Inspection Overhead: 0.99 ms per request
    • Audit Ledger Integrity: chain_valid: true (SHA-256 hash chain verified across all entries)

    Next Steps
    1.

    Route Live Agent Traffic: Point sub-agents through inspect_and_proxy_call using their assigned tenant_id and AgentRole (READ_ONLY_ANALYST, DEVELOPER_AGENT, SWARM_SUPERVISOR, or ADMIN_OPERATOR).
    2.

    Continuous Threat Recall: Query recall_threat_signatures or monitor aegisgateway://telemetry for real-time block rates and latency metrics.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Multi-tenant agent environments are prone to cross-tenant data leaks and indirect prompt injections hidden in tool outputs. This skill provides a zero-trust governance gateway to secure MCP tool calls with cryptographically verifiable audit trails.

    What it does

    • Multi-Tenant Isolation enforces strict namespace boundaries and RBAC to prevent one tenant's agent from accessing another's data.
    • Bi-directional Firewall inspects ingress tool arguments for injections and scrubs egress tool outputs for PII and hidden prompt overrides.
    • Chaos Engineering Harness runs automated red-team evaluations across six adversarial vectors to score system resilience.
    • Tamper-Evident Logging maintains a SHA-256 hash-chained ledger of every decision for immutable compliance auditing.
    • Threat Intelligence Memory utilizes Qdrant to store and recall threat signatures for sub-5ms pattern matching.

    How it works

    1. Register Policies using configure_tenant_policy to define namespaces, allowed roles, and rate limits.
    2. Proxy Calls via inspect_and_proxy_call to wrap every MCP tool interaction in a zero-trust inspection layer.
    3. Verify Integrity by running verify_audit_ledger to mathematically confirm no logs have been altered.
    4. Audit and Recall threat patterns using recall_threat_signatures to refine firewall rules based on historical blocks.

    Frameworks & tools

    Built with Python, FastMCP, and Pydantic v2. Uses Qdrant for vector-based threat memory and SHA-256 for cryptographic integrity.

    Why this beats prompting it yourself

    Prompt-based security is easily bypassed by "ignore previous instructions" attacks hidden in untrusted tool outputs. This skill moves security to the protocol level, using deterministic code and cryptographic verification that an LLM cannot hallucinate or ignore.

    Use cases

    • Multi-tenant SaaS where agents must be strictly confined to specific customer data silos.
    • Financial Compliance requiring an immutable, tamper-evident audit trail of all AI tool executions.
    • Red-Teaming to stress-test new MCP integrations against prompt injection and data exfiltration.

    Known limitations

    Requires Qdrant (can run in-memory) for threat signature recall. Inspection adds a sub-5ms latency overhead per call.

    Runs embedded in-memory Qdrant and an in-memory SHA-256 audit ledger by default (state resets on server restart unless external Qdrant persistence is configured). Upstream MCP forwarding operates in local policy-inspection/simulation mode unless wired to downstream SSE/stdio endpoints.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 2 days ago

    Requires

    Qdrant
    Python 3.11+
    Python

    Permissions

    Terminal / Shell
    Read Files
    Environment Variables

    Allowed Hosts

    qdrant.tech

    File Scopes

    *.json
    *.py
    src/**
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    Standard LLM system prompts are vulnerable to 'jailbreak' attacks hidden in tool outputs; this skill enforces security at the protocol level using deterministic code and cryptographic hashing that an agent cannot ignore or bypass.

    Aegis is built on the Model Context Protocol (MCP) standard, making it compatible with any agent or platform that supports MCP, including Claude Desktop, various IDE plugins, and custom Python/TypeScript implementations.

    The gateway uses a SHA-256 hash-chained ledger to record every tool interaction, ensuring that any attempt to alter or delete logs is mathematically detectable during an integrity check.

    The inspection process is highly optimized, adding less than 5ms of latency per call, which is negligible compared to the standard inference time of most LLMs.

    While Aegis uses Qdrant for high-speed threat signature matching, it supports an in-memory mode for easy setup, though a persistent Qdrant instance is recommended for production environments.

    You receive the full Aegis MCP skill package including the multi-tenant isolation logic, the bi-directional firewall, and the automated chaos engineering harness for red-teaming your setup.

    New in Security