About otx
otx is an MCP server for AI agents researching threat intelligence in AlienVault OTX. It covers indicators, pulses and users, with account-changing tools disabled unless explicitly enabled.
Key features
- 01Enrich indicators across types
- 02Search pulses and users
- 03Retrieve reputation and analysis data
- 04Read-only tools by default
- 05stdio and authenticated HTTP transports
Use cases
- 01Research an IP, domain, hash, URL or CVE
- 02Review community pulses about malware
- 03Inspect passive DNS or sandbox results
MCP support
Exposes AlienVault OTX tools and resource templates to MCP clients. Supports stdio and streamable HTTP; HTTP access can require bearer-token authentication.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
uvx otx-mcpOTX_API_KEYSecretYour AlienVault OTX API keyOTX_MCP_READ_WRITESet to 1 to expose account-mutating tools (create/subscribe/follow/submit). Read-only by default.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project README
Quickstart · Tools · HTTP · Security · Troubleshooting · Docs
mcp-otx is an MCP server that gives AI agents access to AlienVault OTX (Open Threat Exchange, now LevelBlue). Hand your agent an IP, domain, file hash, URL or CVE, or a malware family or adversary to research, and get back what the OTX community knows: pulses, reputation, passive DNS, whois, sandbox analysis and related indicators.
Note
Independent project, not affiliated with AlienVault or LevelBlue. You need your own OTX API key; a free account is enough.
Highlights
- Full API coverage. 29 tools across pulses, indicators, users, subscriptions and URL submissions, plus resource templates for pulses and indicators.
- One-call indicator enrichment. Every OTX indicator type (IPv4, IPv6, domain, hostname, file hash, URL, CVE) and detail section: reputation, geo, passive DNS, malware, whois, URL lists, HTTP scans, sandbox analysis.
- Read-only by default. The 9 tools that change your OTX account only appear with
--read-write. - Context-sized output. Large OTX objects are summarized and capped, with the API's total counts kept, so one lookup doesn't fill the model's context.
- Safe HTTP. stdio for desktop clients, plus streamable HTTP with bearer-token auth. The server refuses to listen on a network-facing address without a token.
- Tested. Unit tests against a mocked OTX API run in CI on Python 3.10 to 3.13.
How it works
flowchart LR
A["AI agent<br/>(Claude, Codex, Gemini, VS Code)"] -->|"MCP over stdio or HTTP"| S["mcp-otx"]
S -->|"HTTPS + X-OTX-API-KEY"| O["OTX DirectConnect API"]
O -->|"JSON"| S
S -->|"summarized, capped JSON"| A
Quickstart
Requirements: Python 3.10 or newer, uv (or pip), and an OTX API key.
# Claude Code
claude mcp add --transport stdio --env OTX_API_KEY=your-key otx -- uvx otx-mcp
# Codex CLI
codex mcp add otx --env OTX_API_KEY=your-key -- uvx otx-mcp
# Gemini CLI
gemini mcp add -e OTX_API_KEY=your-key otx uvx otx-mcp
With pip: pip install otx-mcp, then use the command otx-mcp. For Claude Desktop, VS Code, Cline, Docker and running from source, see Connect a client.
Then ask your agent: "What does OTX know about 8.8.8.8, and are there recent pulses about Emotet?" More prompts are in example workflows.
Tools
20 read tools, plus 9 write tools with --read-write. Full parameters and real sample output are in the tool reference.
| Area | Read tools | Write tools (--read-write) |
|---|---|---|
| Search | search_pulses, search_users | |
| Pulses | get_pulse_details, get_pulse_indicators, get_related_pulses, search_related_pulses, list_pulse_activity, list_subscribed_pulses, get_subscribed_pulse_ids, get_pulse_events | create_pulse, subscribe_pulse, unsubscribe_pulse |
| Indicators | get_indicator_details, validate_indicator, list_indicator_types, export_indicators | |
| Users | get_user, get_current_user, list_user_pulses, list_my_pulses | subscribe_user, unsubscribe_user, follow_user, unfollow_user |
| Submissions | list_submitted_urls, list_submitted_files | submit_url, submit_urls |
Resource templates: otx://pulse/{pulse_id} and otx://indicator/{type}/{value}.
HTTP transport
stdio is the default. For remote or web clients, run streamable HTTP:
OTX_API_KEY=your-key uvx otx-mcp --transport streamable-http
# → http://127.0.0.1:8000/mcp
To listen on a network-facing address, set a bearer token. Clients must then send Authorization: Bearer <token>:
OTX_API_KEY=your-key OTX_MCP_BEARER_TOKEN=a-long-random-secret \
uvx otx-mcp --transport streamable-http --host 0.0.0.0
docker build -t mcp-otx .
docker run --rm -p 127.0.0.1:8000:8000 -e OTX_API_KEY=your-key \
-e OTX_MCP_BEARER_TOKEN=a-long-random-secret \
-e OTX_MCP_TRANSPORT=streamable-http -e OTX_MCP_HOST=0.0.0.0 mcp-otx
OTX is slow on some endpoints (passive DNS and subscriptions can take 20 to 60 s), so give remote clients a long read timeout (120 s works).
Configuration
| Variable | CLI flag | Default | Description |
|---|---|---|---|
OTX_API_KEY | (required) | Your OTX API key. Also read from .env | |
OTX_API_BASE | https://otx.alienvault.com/api/v1 | API root (for testing) | |
OTX_MCP_TRANSPORT | --transport | stdio | stdio, streamable-http or sse |
OTX_MCP_HOST | --host | 127.0.0.1 | HTTP bind address |
OTX_MCP_PORT | --port | 8000 | HTTP port |
OTX_MCP_READ_WRITE | --read-write | off | 1 exposes the account-changing tools |
OTX_MCP_BEARER_TOKEN | off | Require Authorization: Bearer <token> on HTTP | |
--allow-remote | off | Allow a network-facing bind without a token (not recommended) | |
--json-response | off | Plain JSON instead of SSE streams (streamable HTTP) | |
--stateless-http | off | A new session per request (streamable HTTP) |
Security
Warning
The server acts as your OTX account. Keep it read-only unless you need the write tools, and never expose HTTP without a bearer token and TLS.
- Read-only by default. Pulse creation, subscriptions, follows and URL submissions need
--read-write. Editing or deleting pulses isn't exposed as a tool at all. - Untrusted content. Pulse and indicator descriptions are written by the OTX community. The server tells the model to treat them as data, not instructions.
- The API key is sent only to the OTX API, in a header.
Details are in the security model. Report vulnerabilities privately; see SECURITY.md.
Troubleshooting
- 403 "Authentication required": the API key is wrong.
get_current_usershows which account a key belongs to. - A write tool is missing: start the server with
--read-write. - Timeouts: passive DNS and subscription calls can take up to 60 s. Raise your client's tool timeout.
More in the troubleshooting guide.
Development
git clone https://github.com/w0h1v/mcp-otx && cd mcp-otx
uv venv && uv pip install -e ".[dev]"
.venv/bin/python -m pytest # mocked OTX API, no key or network
See CONTRIBUTING.md and the changelog.
License
MIT. See LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.