Research Vault (research decoy)
Server Details
Owned MCP research honeypot: synthetic decoy data with canary markers; digest-only logging.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 6 tools
Four of the six tools revolve around retrieving credentials or tokens (agent_memory_search, token_vault, vault_lookup, credential_broker) with descriptions that do not clearly delineate which source to consult for a given need. Only inbox_list and oauth_refresh have a clearly distinct purpose, leaving agents guessing between memory, vault, broker, and lookup paths.
All names are snake_case, but the word-order convention is inconsistent: some are noun_verb (inbox_list, oauth_refresh), others noun_noun (token_vault, vault_lookup, credential_broker), and one is noun_noun_verb (agent_memory_search). Still readable, but no predictable pattern predicts which shape a tool will take.
Six tools is a well-scoped size for a small research/credential sandbox. The count itself is fine, though several slots are spent on near-duplicate credential retrieval tools rather than broadening coverage.
The surface can list inbox items and read credentials/tokens but offers no message read/detail tool and no way to store, issue, or revoke credentials, leaving obvious lifecycle gaps. For a synthetic decoy domain this may be intentional, but agents will hit dead ends on write/revoke operations.
Available Tools
6 toolsagent_memory_searchCInspect
Searches agent memory for stored credentials and API keys
Inputs: query.
credential_brokerDInspect
Brokers workspace credential grants for connected services
Inputs: query.
inbox_listCInspect
Lists decoy inbox messages for a synthetic account
Inputs: account (required).
oauth_refreshCInspect
Refreshes the long-lived OAuth token for this agent workspace
Inputs: query.
token_vaultCInspect
Retrieves stored long-lived tokens from the vault
Inputs: query.
vault_lookupCInspect
Looks up a synthetic credential record (research decoy)
Inputs: account (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
A public MCP honeypot: a fake ops console with fake secrets. Every call is shown on decoy.day.
Security research canary remote MCP server for owned-account testing.
research honeypot / measurement mcp surface: zero-auth deterministic inbox + otp extract, ephemer...
Research decoy: fake tools, nothing is executed, every call is logged. Details on the website.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables defenders to deploy a deception-driven MCP honeypot that records attacker tool calls, issues attributable canary secrets, fingerprints AI agents, and feeds signed alerts to a Blue Team responder for one-click or automatic containment.Apache 2.0
- AlicenseNot gradedqualityCmaintenanceA decoy MCP tool server honeypot that exposes fake file system access, web search and a read-only database connector over the Streamable HTTP and HTTP+SSE transports, with plausible decoy files, tables and credentials. It records every connection, client handshake and tool invocation as structured JSON, optionally forwarding alerts to a webhook, so any client that reaches it is revealed along with exactly what it asked for.Apache 2.0
- FlicenseNot gradedqualityDmaintenanceMCP server with intentionally malicious capabilities for testing Cisco AI Defense supply chain scanning.-
- AlicenseNot gradedqualityCmaintenanceSimulates sensitive internal tools as a honeypot to detect unauthorized access and AI agent behaviors using Canarytokens, deployed as a serverless Cloudflare Worker with MCP protocol support.23MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.