eosl-mcp

MCP server for EOSL.ai — source-backed hardware end-of-life (EOL / EOSL) lookups by part number. Is this switch/server/firewall still supported? When did — or does — vendor support end? Every answer carries the URL of the manufacturer's own end-of-life bulletin, so nothing is asserted without a source. Unknown parts return found:false, never a guess.

Covers enterprise datacenter gear from Cisco, Dell, HPE, Fortinet, IBM, Juniper, Palo Alto Networks, Arista, and more — current coverage figures live on the site.

Fastest path: the hosted endpoint (no install)

A hosted, no-auth instance runs at https://eosl.ai/mcp (Streamable HTTP), listed in the official MCP registry as ai.eosl/eosl:

claude mcp add --transport http eosl https://eosl.ai/mcp

Related MCP server: Cisco CX Cloud MCP Server

This package: local stdio server

Same five tools, same matching rules, reading the same public data — as a local stdio process.

npx eosl-mcp

Claude Desktop / any stdio client (mcpServers config):

{ "mcpServers": { "eosl": { "command": "npx", "args": ["-y", "eosl-mcp"] } } }

Docker:

docker build -t eosl-mcp . && docker run -i eosl-mcp

Zero dependencies; Node ≥ 18. Verify everything works:

node server.js --selftest

Tools

Matching, honestly

Exact match first, then punctuation-insensitive, then vendor-gated Fortinet short-SKU aliases (FG-60E → FortiGate-60E) — never fuzzy. If a caller names a vendor, a match from any other vendor is rejected: another vendor's dates are worse than no answer.

Data and privacy

• Data source: https://eosl.ai/data/lookup.json — the same open dataset behind the site, CC BY 4.0, refreshed weekly from vendors' own published notices.

• This local server sends only the HTTP fetches above; part numbers you look up locally are matched in-process against the downloaded dataset copy for lookup_part/bulk_check/search_models (only get_family fetches per-slug). The hosted endpoint records aggregate usage as described at eosl.ai/api/usage.

• Always confirm critical dates against the linked vendor bulletin before acting on them.

License

Code: Apache-2.0. Dataset: CC BY 4.0 — attribution to EOSL.ai.