Precursor Security
Security & Access

Read-only by design.

The Precursor Intelligence MCP server is read-only. All 41 tools query intelligence and return data; none can write to your systems, modify your data, or trigger scans. Access uses a bearer API key that is scoped, IP-allowlisted, logged on every request and revocable at any time.

Start freeBack to overview

What the MCP can and cannot do

A useful way to reason about any agent tool is the blast radius. Here the radius is zero: the server can read intelligence, and nothing else.

It can

  • Look up any CVE with EPSS, CVSS and CISA KEV status
  • Return Shadowserver honeypot activity for a CVE or vendor
  • Read MITRE ATT&CK techniques, threat actors and IOCs
  • Return AI-generated remediation summaries that already exist

It cannot

  • Write, edit or delete any of your data
  • Trigger network or attack surface scans
  • Reach into your code, repositories or infrastructure
  • Change triage, watchlists or platform settings

The access model

Every connection runs over Streamable HTTP, authenticated with a key you control end to end.

Bearer key authentication
Every request carries a bearer API key minted in the dashboard. The secret is shown once and stored only as a hash, so plaintext never rests on our side.
Scoped and IP-allowlisted
Keys are issued with read scope for the MCP, and can be pinned to a set of allowlisted IP addresses so a leaked key is not usable from anywhere.
Rate limited
A default 60 requests per minute per key. The limit, remaining count and reset window come back on every response, so clients can back off gracefully.
Logged on every call
Each request is logged with your organisation, key, endpoint, status and latency, giving you an attributable trail of what your agents queried.
Revocable in seconds
Revoke a key from the dashboard and it stops working within seconds. Keys can also carry an expiry date.
Metered, not open-ended
Usage is metered per tool call, so agent activity stays predictable and attributable rather than an unbounded firehose.

Manage your keys

Mint, allowlist and revoke keys in the dashboard under Settings, then API Keys.

Open API Keys

Try it with a free key

Read-only access, revocable in seconds. Connect your agent and see exactly what it can pull.

Start freeView setup

Security questions, answered

What the Precursor Intelligence MCP can and cannot do.

No. All 41 tools are read-only queries. The server returns intelligence such as CVE, EPSS, KEV and honeypot data. It has no tools that write, delete or change anything.

No. The MCP server does not scan. It cannot reach into your environment. Attack surface scanning exists in the platform dashboard as a separate, deliberate action and is not exposed through the MCP.

With a bearer API key you mint in the dashboard under Settings, then API Keys. It is passed in the Authorization header. There is no OAuth and no password exchange over the protocol.

Revoke it in the dashboard. Revocation takes effect within seconds. Since only a hash of the secret is stored, you mint a fresh key to restore access.

Yes. Keys have a default limit of 60 requests per minute. The limit, remaining count and reset window are returned with every response so a client can back off cleanly.