OpenClaw

OpenClaw is an open-source personal AI agent that runs on your own machine and speaks the Model Context Protocol natively. Point it at the PostFuze MCP server and your agent gains all 32 PostFuze tools — it can connect social accounts, draft and schedule posts, upload media and read analytics from plain-English instructions, with no glue code.

Because OpenClaw is MCP-native, there is no bridge or proxy to install: it connects to PostFuze over Streamable HTTP and authenticates with the very same sk_live_… key as the REST API. The key scopes every action to your organization, so the agent can only touch accounts your key can reach.

How a request flows

OpenClaw turns your instruction into one or more tools/callrequests. Each tool is a thin dispatch over PostFuze’s REST API, which fans a post out to every connected platform and reports per-account status back up the chain.

Add PostFuze to OpenClaw

OpenClaw reads MCP servers from an mcpServers map in its openclaw.json config (in your OpenClaw config directory, typically ~/.config/openclaw/openclaw.json). Add a postfuze entry pointing at the Streamable HTTP endpoint:

{
  "mcpServers": {
    "postfuze": {
      "type": "http",
      "url": "https://api.postfuze.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk_live_…"
      }
    }
  }
}

Prefer to keep the secret out of the file? OpenClaw expands ${ENV_VAR} references, so you can store the key in the environment and reference it instead:

{
  "mcpServers": {
    "postfuze": {
      "type": "http",
      "url": "https://api.postfuze.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${POSTFUZE_API_KEY}"
      }
    }
  }
}

# Export your key, then start OpenClaw so it picks up the server.
export POSTFUZE_API_KEY="sk_live_…"
openclaw run

What it unlocks

Once connected, the agent can chain these PostFuze tools on your behalf. The full list is on the tool reference:

• Publish & schedule — create_post (with first_comment and thread_content), list_posts, get_post, delete_post.
• Accounts — list_social_accounts, get_social_account, get_account_metrics, and get_auth_url to connect a new platform.
• Media — upload_media → confirm_media_upload to attach images and video (for video, wait for ready first; pick a cover with set_media_thumbnail).
• Connections & webhooks — get_account_health to verify a channel before posting, list_pending_pages / finalize_connection to finish a multi-account OAuth connection, and register_webhook / test_webhook to wire up event delivery.
• Analytics & usage — get_post_analytics and get_account_usage to report on performance and quota.

Example prompts

Talk to OpenClaw the way you would a teammate. It picks the right tools, asks for anything missing, and reports back:

"Draft a post for my X and LinkedIn accounts announcing our v2 launch,
 add 'Read the changelog: https://acme.com/v2' as the first comment, and
 schedule it for 9am tomorrow in my timezone."

"Which of my connected accounts gained the most followers in the last 30 days?"

"Pull the engagement metrics for my most recent published post and tell me
 which platform performed best."

For the first prompt OpenClaw will call list_social_accounts to resolve the account IDs, then create_post with scheduled_at set and first_comment populated. For the analytics prompts it reaches for get_account_metrics and get_post_analytics respectively. To attach a video, it chains upload_media → confirm_media_upload, waits for the asset to reach ready, then calls create_post; platforms that need an extra field (e.g. a Pinterest board) are handled via the post configuration. See publishing a video with an agent and per-platform configuration.

Next steps

• Tool reference — every tool, its parameters and the two resources.
• Claude Desktop and Cursor & Windsurf — other MCP clients.