Package Information
Available Nodes
Documentation generating…
Documentation
n8n-nodes-agentnexus
n8n community node for Agent Nexus: a registry of external APIs, MCP servers and CLIs, each one health-checked continuously. The registry answers a plain-language need ("send a transactional email") with the callable interfaces that match, each carrying live reliability signals.
Operations
- Discover: describe a need in plain language and get matching callable interfaces with reliability signals attached.
- Get Entry: full details of one registry entry by slug.
- Get Status: public uptime and health-check totals.
All three are usable as tools inside n8n AI Agents (usableAsTool), so an agent can pick an external tool itself, call it, and read the result.
Credentials
Optional. The node uses the AgentNexusApi credential:
| Field | Value |
|---|---|
| API Key | A free Agent Nexus key, e.g. nx_live_.... Leave empty for anonymous access. |
- Without a key: 100 calls/day per address.
- With a free key: 1,000 calls/day. Get one without an account:
curl -X POST https://agentnexus.app/api/public/keysor visit agentnexus.app/free-key.
The key is sent as an x-api-key header on every request.
Worked example: Discover
Step 1. Add the node and pick the Discover operation.
Step 2. Type the need in plain language in the Need field:
send a transactional email
Optional parameters:
- Category:
API,MCP ServerorCLIto restrict the layer. Leave asAny. - Minimum Reliability:
0-100. For example60keeps only interfaces scoring 60 or above. - Limit: maximum results to return (1-50).
Step 3. Run it. The node sends GET https://agentnexus.app/api/public/discover?need=send%20a%20transactional%20email&limit=5 and returns JSON like this (trimmed to one match):
{
"need": "send a transactional email",
"coverage": "exact",
"count": 1,
"uncovered": false,
"note": "Matched on declared capabilities. Call the interface with the `call` contract, then report the outcome.",
"matches": [
{
"slug": "smtp2go-api",
"name": "SMTP2GO",
"category": "api",
"summary": "Transactional email over a JSON endpoint or plain SMTP relay.",
"match_score": 31.95,
"matched_on": "capability",
"capabilities": ["send email", "transactional email", "smtp relay"],
"call": {
"endpoint": "https://api.smtp2go.com/v3/email/send",
"auth_mode": "API key header",
"input_format": "application/json",
"output_format": "application/json",
"rate_limit": null,
"pricing": "1k emails/month free",
"example": null,
"docs_url": "https://apidoc.smtp2go.com/documentation/"
},
"trust": {
"verified": false,
"reliability_score": 78,
"uptime": 1,
"samples": 9,
"agent_reports": 0,
"avg_latency_ms": 1185,
"last_probe_ok": true,
"last_probe_at": "2026-10-01T04:17:14.349+00:00"
},
"report": {
"endpoint": "https://agentnexus.app/api/public/report",
"method": "POST"
}
}
],
"next_steps": ["..."]
}
Step 4. Use the result downstream. Typical pattern:
- A downstream HTTP Request node reads
matches[0].call.endpointas the URL andmatches[0].call.input_formatas the content type. - Authentication: the provider's own credential (the node returns
call.auth_modeso you know what to configure). The Agent Nexus key is only for the registry, never for the provider's API. - Optionally a final HTTP Request reports the outcome to
POST /api/public/reportwith{"slug":"smtp2go-api","outcome":"success","latency_ms":412}. Reported outcomes feed the reliability scores other agents see, and raise your daily quota by up to +250/day.
In an AI Agent workflow, expose the node as a tool and ask the agent in plain language ("find a tool to send an email"): it calls Discover, reads trust.reliability_score before choosing, and can fall back to the next match when a call fails.
What each response field means
Top level:
| Field | Meaning |
|---|---|
need |
The normalized need as understood by the registry. |
coverage |
exact when declared capabilities match the need, partial when only some tokens matched, none when nothing matched. |
count |
Number of matches returned. |
uncovered |
true when the need is a coverage gap: matches then lists the most reliable interfaces in that layer as starting points only, never as a match. |
note |
One-line explanation of how the result was built. |
next_steps |
Suggested follow-up calls (e.g. get_entry for the top match). |
Each item in matches:
| Field | Meaning |
|---|---|
slug |
Stable registry ID. Use it for Get Entry and for outcome reports. |
name, category, summary |
Display name, interface layer (api, mcp or cli) and one-line description. |
match_score |
Ranking score (higher is better) computed from capability, reliability and popularity signals. |
matched_on |
capability when the match comes from declared capabilities, fallback when it is a starting-point suggestion. |
capabilities |
The declared capabilities that drove the match. |
call.endpoint |
The real URL to call. |
call.auth_mode |
How the provider authenticates (e.g. API key header, Bearer token, none). |
call.auth_params |
Parameter names the provider expects for authentication, when declared. |
call.input_format / call.output_format |
Expected request and response formats (application/json, text/plain, ...). |
call.rate_limit |
Provider-side rate limit as declared by the publisher. |
call.pricing |
Pricing summary as declared by the publisher. |
call.example |
Ready-to-run invocation example, when the publisher supplied one. |
call.docs_url |
Provider documentation link. |
trust.verified |
Publisher identity verified by Agent Nexus. |
trust.reliability_score |
0-100 score from Agent Nexus health probes and agent outcome reports. |
trust.uptime |
Share of probes answered OK (0-1) over the tracked window. |
trust.samples |
Number of probes behind the score. |
trust.agent_reports |
Number of real outcome reports from calling agents. |
trust.avg_latency_ms |
Average round-trip latency in milliseconds. |
trust.last_probe_ok / trust.last_probe_at |
Result and time of the most recent health probe (ISO 8601, UTC). |
report |
Ready-made outcome-report payload for closing the loop after a real call. |
Reliability signals come from Agent Nexus health probes (liveness, capability and response-schema checks, run continuously and published at agentnexus.app/status) plus outcome reports from calling agents. They describe the interface, not your workflow.
Other operations
- Get Entry takes a
slug(e.g.smtp2go-api) and returns the complete registry record: endpoint, authentication parameters, formats, rate limits, pricing, docs link and health history. Use it once a candidate is chosen, to build the actual request. - Get Status returns the registry's public totals: entries monitored, checks run, success rate, average latency. Use it to display registry health inside your workflow.
Install
Settings > Community Nodes > Install > n8n-nodes-agentnexus
License
MIT