Precursor Security
Precursor Intelligence MCP

Threat intelligence
for the AI era.

Your scanner finds thousands of issues. Now your agent can ask which ones actually matter. 41 read-only intelligence tools for Claude Code, Cursor and any MCP client.

The Precursor Intelligence MCP server is a read-only threat intelligence server for AI agents. It exposes 41 tools over the Model Context Protocol (Streamable HTTP), so tools like Claude Code, Claude Desktop, Cursor and Zed can query CVE, EPSS, CISA KEV, MITRE ATT&CK, IOC and Shadowserver honeypot data and answer "which vulnerabilities actually matter to my stack?".

Free tier. Read-only. No agents to deploy.
41 tools
read-only intelligence tools across 11 categories
Streamable HTTP
Model Context Protocol transport (spec 2025-06-18), one bearer token
2 to 5%
of published CVEs are ever exploited, and the agent finds them
3,000+ Assessments DeliveredTriple-CREST Accredited24/7 UK SOC in NewcastleReports Accepted by Insurers & RegulatorsEst. 2018
Up and running in minutes

Install it, ask a question, get an answer

One command adds Precursor Intelligence to your agent. From then on, security context is a question away.

01
Get a key
Sign up free and mint an API key from Settings, then API Keys. No sales call, no card.
02
Add the server
One command in Claude Code, or a short block of config in Cursor or Claude Desktop.
03
Ask your agent
Ask which CVEs matter, which are exploited, and what to fix first. Answers come back with evidence.
claude mcp add --transport http precursor-intelligence \
  https://mcp.precursorintelligence.com \
  --header "Authorization: Bearer <YOUR_API_KEY>"
41 tools, 11 categories

One server, the whole intelligence graph

Every dataset the platform curates, queryable by your agent through a single connection.

CVE, EPSS and CISA KEV5 tools

Full intelligence on any CVE, including EPSS probability, KEV status and honeypot activity.

get_cvesearch_cves_by_vendorlist_high_epss_cveslist_kev_recent
Vendors and products (CPE)5 tools

Map CVEs to specific vendors, products and versions across the CPE dictionary.

get_vendorget_vendor_product_versionsearch_products
Weaknesses (CWE)3 tools

Look up the underlying weakness classes behind vulnerabilities.

get_cwelist_cweslist_top_cwes
MITRE ATT&CK4 tools

Techniques, tactics and sub-techniques with detection and kill-chain context.

get_mitre_techniquelist_mitre_tacticslist_mitre_techniques
Atomic Red Team3 tools

Atomic tests mapped to ATT&CK techniques for detection validation.

list_art_techniquesget_art_atomic_test
Threat actors3 tools

Groups by aliases, sponsor, target industry and incident type.

get_threat_actorlist_threat_actors_by_industry
Indicators of compromise3 tools

Search and enrich IOCs across the ThreatFox feed and blog-derived indicators.

search_iocslookup_ioclist_recent_iocs
Malware samples4 tools

MalwareBazaar samples by hash, file type or signature.

get_malware_samplesearch_malware_sampleslist_top_malware
C2 and TLS blacklists2 tools

Check SSL certificate and JA3 fingerprints against abuse.ch blacklists.

lookup_ssl_blacklistlookup_ja3_blacklist
Shadowserver honeypot4 tools

Live honeypot exploitation activity per CVE, vendor and day.

get_shadowserver_cve_activitylist_shadowserver_top_cves
Remediation and account5 tools

AI-generated remediation summaries, the remediation queue, data freshness and account status.

list_remediation_summarydata_freshnessaccount_status
What you can ask

Stop reading advisories. Start asking questions.

Which CVEs affecting Fortinet are actively being exploited right now?

Give me the highest-EPSS vulnerabilities added to CISA KEV in the last 14 days.

Write a remediation plan for CVE-2025-53770.

Map CVE-2024-3400 to the MITRE ATT&CK techniques it enables.

Which threat actors target healthcare, and what do they use?

Is this SHA1 certificate fingerprint linked to known C2 infrastructure?

Built for enterprise trust

Read-only by design

All 41 tools are read-only. The server returns intelligence, it cannot modify your data, write to your systems or trigger scans. Access is a bearer key you mint and revoke yourself.

Read the security model
Read-only tools
No writes, no scans, no changes to your estate.
Bearer key auth
Minted in the dashboard, scoped and IP-allowlisted.
Revoke any time
Kill a key and access stops within seconds.
Every call logged
Org, endpoint, status and latency, attributable.
Who it is for

Security context, where the work already happens

AI-native developers
You already build with Claude Code, Cursor or Zed. Give your coding agent real vulnerability context before it recommends a fix or bumps a dependency.
Lean security teams
Too many CVEs, too few people. Turn a weekly flood of advisories into a short, evidenced action list your agent can draft for you.
MSPs and consultants
A lightweight intelligence layer you can query per client, with no agents to deploy and no scanning of client estates.

Give your agent real security context

Mint a free key, add the server, and ask your first question in under two minutes. Explore the full platform or read how the exploitation risk score works.

Already have a key? Log in.

The MCP server, explained

How the Precursor Intelligence MCP works, and what it can and cannot do.

It is a read-only threat intelligence server for AI agents. It exposes 41 tools over the Model Context Protocol, so an agent like Claude Code or Cursor can pull CVE, EPSS, CISA KEV, MITRE ATT&CK, IOC, malware and Shadowserver honeypot data directly into your workflow and tell you which vulnerabilities actually matter.

It works with Claude Code, Claude Desktop, Cursor, Zed, and any MCP-compatible client. It uses the Streamable HTTP transport, so you point the client at the server URL and add your API key as a bearer token.

No. Every one of the 41 tools is read-only. The server can look up and return intelligence, but it cannot modify your data, write to your systems, or trigger network scans. Attack surface scanning is a separate dashboard feature and is not exposed over the MCP.

With a bearer API key generated in the platform under Settings, then API Keys. You pass it in the Authorization header. There is no OAuth and no password exchange, and a key can be revoked at any time.

Yes, there is a free tier. You mint an API key from the dashboard and start querying. Usage is metered per tool call, and lightweight calls such as checking your account status are free.

The public intelligence datasets Precursor Intelligence curates: CVEs with EPSS and CISA KEV status, CWE weaknesses, MITRE ATT&CK techniques, threat actors, IOCs, malware samples, C2 blacklists and Shadowserver honeypot activity. It does not access your internal systems or code.