
Abnormal MCP Server
io.github.GregDogPublished by GitHub account GregDogv1.1.0Updated Oct 24, 2026
Abnormal Security MCP: threats, search, remediation, ATO cases, vendor/BEC, and evidence download.
Overview
Lets an assistant query Abnormal Security threat, message, employee, ATO case, and vendor data, with optional remediation and evidence download tools.
- What you can do
- An MCP server for Abnormal Security that exposes 27 always-on read tools covering the threat log, message search, mailbox campaigns, employee identity and logins, ATO cases, vendor compromise cases, Detection 360 reports, URL rewrite clicks, and audit logs. Optional opt-in tool sets add response actions (search remediation, threat remediate/unremediate, case status update, Detection 360 report submit) and evidence download (message EML and attachments). It runs as a local process over stdio by default, with opt-in Streamable HTTP.
- When to use it
- Use it when an assistant needs to investigate email threats, business email compromise, account takeover cases, or vendor risk from Abnormal Security, and when read-only access is enough. Enable the response or evidence gates only if you actually need remediation or message/attachment retrieval.
- Requirements
- A local runtime (native Go binary or the Docker image ghcr.io/gregdog/mcp-server-abnormal). ABNORMAL_API_TOKEN is required; ABNORMAL_BASE_URL selects the US, EU, or FedRAMP endpoint. Optional gates: ABNORMAL_ALLOW_RESPONSE, ABNORMAL_ALLOW_EVIDENCE_DOWNLOAD, ABNORMAL_MAX_EVIDENCE_BYTES. Network access to the Abnormal REST API.
Installation
In SourceWeft
- Open Abnormal MCP Server in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
Abnormal MCP Server
A secure, open-source Model Context Protocol server for Abnormal Security.
This project is not an official Abnormal product and is not endorsed by Abnormal AI.
Overview
abnormal-mcp lets MCP clients such as Cursor and Claude Desktop query Abnormal Security over stdio (default) or Streamable HTTP. It uses a lightweight handwritten client against the Abnormal Security Client API.
The server is read-only by default. Response and evidence download tools are opt-in and independent. Abnormal authorization still applies to every request.
Features
Read (always on)
- Threat list and get from the Threat Log
- Threat links and attachment metadata
- Threat action status poll (
abnormal_threat_action_get) - Message search with ergonomic filters (
since,sender,sender_domain,recipient,subject,url,attachment,sender_ip,judgement) - Search activity list and status
- Message remediation history
- AI Security Mailbox (formerly Abuse Mailbox) campaigns and unanalyzed reports
- Employee profile, identity (Genome), and recent logins
- ATO case list, detail, and analysis
- Vendor list, detail, activity, and vendor compromise cases
- Detection 360 report list (false positives and missed attacks)
- US, EU, and FedRAMP base URLs (configurable)
Response (opt-in: ABNORMAL_ALLOW_RESPONSE=true)
- Search remediation (
delete,move_to_inbox) withconfirm: truepreview gate - Threat remediate / unremediate with
confirm: truepreview gate - ATO case status update with
confirm: truepreview gate - Detection 360 report submit (false positive, missed attack/spam/graymail) with
confirm: truepreview gate
Evidence download (opt-in: ABNORMAL_ALLOW_EVIDENCE_DOWNLOAD=true)
-
Message EML download (by ABX message ID or search
cloud_message_id) -
Attachment analysis signals and attachment download
-
Metadata + bounded preview by default; optional base64 embed (max 1 MiB in tool output)
-
stdio transport (default) and opt-in Streamable HTTP
-
Native Go binary and Docker image
-
MCP Registry listing on tagged releases (
io.github.GregDog/mcp-server-abnormal)
Quick Start
Create an Abnormal REST API token in the Abnormal portal under Integrations → Abnormal REST API.
Installation
From source
Docker
Client configuration
Cursor (stdio)
Claude Desktop
See examples/claude-desktop.json.
Tools
27 read tools are always registered. With ABNORMAL_ALLOW_EVIDENCE_DOWNLOAD=true, five evidence tools are added (32 total). With ABNORMAL_ALLOW_RESPONSE=true, four response tools are added (31 total, or 36 with both gates enabled).
See docs/tools.md for parameters.
Configuration
Full list: docs/configuration.md.
Security
Read tools are always on. Response and evidence tools require explicit opt-in.
Local use only: this server has no built-in authentication or RBAC. Use stdio (default) or loopback HTTP on the same machine. Do not expose the HTTP endpoint on a network without your own access controls.
See docs/security.md and SECURITY.md.
Development
See docs/development.md.
License
Apache-2.0. See LICENSE.
Source: README.md at commit 4f7bfce
Tools
0Version history
2- v1.1.0LatestSep 17, 2026
- v1.0.1Sep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts