AI-readable Zevari documentation snapshot

Hermes Agent Integration: Connect to LinkedIn via Zevari

Give a Hermes Agent safe LinkedIn access through Zevari - register the MCP server in config.yaml or call the REST API, and run an approval-gated find, score, stage, approve loop.

Requested URL: https://docs.zevari.ai/api/connect/hermes-agent

Overview

Hermes Agent, by Nous Research, is a self-hosted autonomous agent framework with persistent memory and a self-improving loop, VPS-hosted, configured at /opt/data/config.yaml, with native MCP support. This is Hermes Agent the framework, not the separately named open-weight LLM family. Zevari is the LinkedIn execution layer it calls.

Prerequisites

A Zevari workspace with an active subscription, an API key (zvr_api_..., sent as Authorization: Bearer), and a connected LinkedIn sender. The canonical REST reference is the bring-your-own-agent page.

Connect via config.yaml (MCP)

Hermes Agent supports MCP natively. Add Zevari to the mcp_servers block in /opt/data/config.yaml with url https://mcp.zevari.ai/mcp and auth Bearer ${ZEVARI_API_KEY}; the agent then discovers Zevari's LinkedIn tools automatically.

Connect via REST (bearer)

For the REST alternative, call https://api.zevari.ai with Authorization: Bearer <key> from the agent's own code.

A working loop: find, score, stage, approve

Search posts or profiles, ICP-score the matches, draft in your voice, create a campaign, then stage each write with confirmations.requestAction. A human approves, and only then does the write execute. A persistent, self-improving Hermes Agent still never sends unattended.

Safety and approval notes

Every LinkedIn write is staged for approval. Server-side weekly ceilings (Free 40, Premium 80, Sales Navigator 150), working hours, and burst caps are enforced, and no browser cookies are used.