n8n-nodes-speko

Give your n8n workflows a phone. This community node lets Speko place outbound voice-agent calls, read back transcripts and recordings, and convert speech in either direction — all from one credential.

n8n is a fair-code licensed workflow automation platform.

Installation

Follow the community nodes installation guide, then enter n8n-nodes-speko as the npm package name.

Self-hosted, from the command line:

npm install n8n-nodes-speko

Credentials

Create an API key at platform.speko.dev/api-keys and paste it into a Speko API credential. Keys start with sk_live_. The credential test calls GET /v1/organization, so a green check means the key can actually reach your workspace.

Operations

Speko

| Resource | Operation | What it does | | --- | --- | --- | | Call | Place | Calls a phone number and runs one of your voice agents on the line | | Call | Get | Looks up a call — status, duration, and outcome | | Call | Get Transcript | Returns the turns plus a flattened text field of the whole conversation | | Call | Get Recording | Downloads the call audio into a binary field | | Agent | Get Many | Lists the voice agents in your workspace | | Agent | Get | Fetches one agent by ID | | Speech | Synthesize | Turns text into speech audio in any supported language | | Speech | Transcribe | Turns an audio file into text |

The Agent field is a dropdown loaded live from your workspace, so you pick a name instead of pasting an agent_… ID.

Prompt variables. If your agent prompt contains {{placeholders}}, fill them per call under Additional Fields → Prompt Variables. Leave it empty and the prompt is sent as written.

AI Agent tool. The Speko node is marked usableAsTool, so an n8n AI Agent can call it directly — that is the "give your AI agent a phone" path with no extra wiring.

Speko Trigger

Starts a workflow when a Speko call event arrives. This is a real subscription, not a poll: the node registers a workspace webhook when the workflow activates and deletes it when the workflow is deactivated.

Events: call.pre_call, call.status, call.report, call.analysis, call.recording.

call.report is the one to reach for by default — it is durable, so a failed delivery is retried on a backoff rather than dropped.

Leave Agent Names or IDs empty to fire for every agent, or select specific agents to narrow it.

Compatibility

Requires n8n 1.82 or later and Node.js 20.15 or later.

Example: after-hours callback

1. Webhook — a form on your site posts a lead.
2. Speko → Call → Place — call the lead with your Front Desk agent, passing {{name}} as a prompt variable.
3. Speko Trigger → Call Report — in a second workflow, catch the finished call.
4. Speko → Call → Get Transcript — pull the conversation.
5. Write the outcome into your CRM.

Verifying against a live workspace

npm run build && node scripts/live-verify.mjs exercises every path the node uses against a real Speko workspace: credential test, agent list and lookup, a synthesize → transcribe round trip through the node's own SSE parser, the outbound-call body contract (without dialing anyone), and three webhook create/check/delete cycles that must leave nothing behind.

It reads an sk_live_ key from files already on the machine and never prints it. Keys are environment-scoped — a staging key returns 401 against api.speko.dev — so the script probes each candidate and reports which environment it landed in.

Note on audio format: /v1/synthesize has no output-format parameter. The format follows whichever provider the router picked, so the same workflow can receive audio/mpeg on one run and audio/pcm;rate=24000 on the next. Raw PCM is headerless and plays nowhere, so the node wraps it in a RIFF header before it reaches the binary field. The audio in your workflow is always openable.

Resources

• Speko documentation
• n8n community nodes documentation

License

MIT