GlobalHack Map

For agents

Put Global Hack Map inside your agent

Read the situation, check your stack and turn a CVE into a pull request. Connect over MCP, or give your coding agent a fix runbook.

https://globalhackmap.vercel.app/api/mcp
Read llms.txt
  • No Global Hack Map API key
  • Six read-only tools
  • No paid research calls

Connect over MCP

Claude Code

Run this command in your terminal.

Terminal
claude mcp add --transport http globalhackmap https://globalhackmap.vercel.app/api/mcp

Using another client? Add a remote Streamable HTTP MCP server with the Global Hack Map URL. No API key is needed.

Tools

Six tools, one sourced picture

Results preserve confirmed, reported and claimed statuses. Cite the source links and the status together. A claim is an attacker’s assertion.

get_situation()
The cached picture: important incidents, trending and exploited CVEs, malicious packages and active outages.
patch_now()
What attackers are exploiting today, per product: what they can do, who is affected and the fix.
search_incidents(query?, country?, status?, since_days?, limit?)
Find incidents by name, actor or domain, with country, status and date filters.
get_incident(id)
A full incident record, its sources, status and practical guidance.
get_cve(cve)
Merged CVE facts, a runbook URL and the exact prompt for a repository fix.
check_stack(input)
Check a manifest, lockfile or product list against public vulnerability data.
Every result keeps its statusConfirmedReportedClaimed

For stack checks, remove credentials, private registry URLs and other sensitive information before sharing input. No result is a guarantee that your stack is safe.

Limits and troubleshooting

Use Streamable HTTP at the exact server URL above. Stack checks accept up to 500 KB and check at most 1,000 packages; read the coverage notes in each result.

Requests are limited per IP. On HTTP 429, wait for the Retry-After interval before retrying. HTTP 503 means a service or request protection is temporarily unavailable.

Confirmed, reported, claimed, precautionary and disputed are distinct statuses. An empty result means no matching record in the available data, not proof of safety.

Runbooks

Fix a CVE from your coding agent

Point your coding agent at /fix/<CVE-ID>. Each runbook:

  1. 01Checks whether the repository is affected
  2. 02Lists the available version evidence
  3. 03Guides a minimal fix
  4. 04Asks for tests before a pull request

Trending now

CVE-2026-111497pip brotli

Read the runbook
Prompt for your coding agent
Check whether this repository is affected by CVE-2026-111497 and fix it if it is.

Fetch https://globalhackmap.vercel.app/fix/CVE-2026-111497 (use curl if you have no fetch tool) and follow the runbook. Find every place the affected package or product is used, upgrade to a fixed version, run the tests, and open a pull request that explains the change. If the repository is not affected, say so and explain why. Ask me before running anything that costs money.

The agent opens with this prompt and nothing runs until you send it. It must verify the vendor advisory, preserve your work and stop if the evidence is insufficient. Runbooks never authorise exploit code, paid actions, deployment or merging.

JSON API

Read the snapshot

The same cached picture, as JSON. Check its generation time and feed availability alongside the records.

Fetch the snapshot
curl -fsSL 'https://globalhackmap.vercel.app/api/snapshot'
Open snapshot JSON

Go deeper

Research with Valyu

Global Hack Map is built on Valyu’s search and DeepResearch APIs. For cited research beyond the snapshot, agents can use Valyu’s own MCP server or the valyu CLI. Ask before research that costs money.

Valyu MCP server URL
https://mcp.valyu.ai/mcp
Valyu documentation