Give an AI agent a practical way to work with your FeedHive content. These tools connect OpenClaw, Claude Code, and command-line workflows to the FeedHive public API. List connected social accounts, prepare posts, organize content, fill plan slots, and read available analytics—all through the same API you can use from a script.
Start with a read-only request, then choose what to create or change. Your agent uses your FeedHive API key and the permissions of your account; it does not connect social accounts for you or bypass FeedHive's approval flow.
| Where you work | Install or run | What you get |
|---|---|---|
| OpenClaw | npx @feedhive/setup-openclaw |
A FeedHive skill with bundled API scripts and documentation in your OpenClaw workspace. |
| Claude Code | npx @feedhive/setup-claude-code |
A FeedHive social skill with the same API scripts and documentation in your Claude Code skills directory. |
| Terminal, CI, or another agent that runs commands | npx @feedhive/cli --help |
JSON-first feedhive commands for the public API, without either agent setup package. |
| Any MCP-compatible client | Connect to https://mcp.feedhive.com |
FeedHive's hosted MCP server: the public API as MCP tools, nothing to install. |
The three npm packages are built from this repository. The setup packages install skills, and the CLI is a separate package. The hosted MCP server is run by FeedHive and needs no package.
With Node.js 20+, run this from the OpenClaw workspace you want to configure, using the API key from your FeedHive account:
npx @feedhive/setup-openclaw <your-api-key>Setup validates your key, installs the FeedHive skill and its scripts, and stores the key in that workspace's .env.local. Then try asking your agent: “List my connected FeedHive accounts and show me my recent posts.” Review the returned data before asking it to create or change content. Explore the OpenClaw skill.
With the same prerequisites, run:
npx @feedhive/setup-claude-code <your-api-key>Setup validates your key, installs the FeedHive social skill under ~/.claude/skills/feedhive, and stores the key in ~/.feedhive/agent-tools.env. Try asking Claude Code: “Show me my FeedHive drafts and connected social accounts.” Explore the Claude Code skill.
If your agent can execute terminal commands, it can use the CLI and read its JSON output. No native plugin is implied:
npx @feedhive/cli socials list
npx @feedhive/cli posts list --limit 20
npx @feedhive/cli analytics social <social-id>For a permanent CLI installation, run npm install -g @feedhive/cli and use feedhive in place of npx @feedhive/cli. See the CLI command reference for arguments, JSON bodies, output, and exit codes.
FeedHive runs an official remote MCP server at https://mcp.feedhive.com (Streamable HTTP). Add it to your MCP client as an HTTP server and send your API key as an Authorization: Bearer YOUR_FEEDHIVE_API_KEY header. Keep the key in the client's private credential settings, not in a chat or a URL.
Tools are named feedhive_{resource}_{action} and cover posts, labels, media, plan slots, social accounts, analytics, and API status. Your configured automation triggers also appear as trigger_{id} tools. Start with a read-only call such as feedhive_socials_list. See the MCP documentation for setup and the full tool list.
| Workflow | Public API operations available here |
|---|---|
| Find your content | List, fetch, create, update, and delete posts; filter post lists by status, labels, and social accounts. |
| Organize a campaign | List and manage labels; attach label IDs when creating or updating posts. |
| Prepare media posts | Create and complete media uploads; list and inspect media; reference completed media IDs in posts. |
| Plan a publishing queue | List and manage plan slots; assign eligible posts to the next slots, or schedule a post for a future time. |
| Inspect your channels | List or fetch connected social accounts accessible to your API key. |
| Check performance | Read post-level and connected-account analytics where the provider supplies them. |
These are API-backed operations, not autonomous content generation or guaranteed publishing. Scheduling needs a connected account and a valid future time; where your FeedHive workspace requires approval, a scheduled API post or assigned post remains pending approval until approved in FeedHive. Analytics fields vary by social provider, and a response marked stale: true can be a usable cached result. See the public API reference for request shapes and limitations.
First inspect the accounts and existing posts:
npx @feedhive/cli socials list
npx @feedhive/cli posts list --status draftCreate a draft by saving this JSON as post.json:
{
"text": "A first draft for our next update",
"status": "draft"
}npx @feedhive/cli posts create --body-file ./post.jsonTo schedule a post, use an account ID returned by socials list and an ISO 8601 time in the future in your JSON body:
{
"text": "Our next update",
"accounts": ["<your-social-account-id>"],
"status": "scheduled",
"scheduled_at": "<future-ISO-8601-datetime>"
}Save the body as scheduled-post.json, then use the ID returned when you created the draft:
npx @feedhive/cli posts update <post-id> --body-file ./scheduled-post.jsonYou can also create a new scheduled post with posts create --body-file ./scheduled-post.json. Check the returned status: approval rules may leave it pending rather than ready to publish. Use plan slots when you want FeedHive to place eligible posts into your queue instead of choosing a date yourself. Write commands affect your real account; inspect the response and confirm the intended accounts and schedule.
Get a public API key from your FeedHive account and pass it to the installer as shown above. If you leave the argument out, the installers and the CLI read FEEDHIVE_API_KEY from your environment. The CLI also reads the installer credential file at ~/.feedhive/agent-tools.env and the current workspace's .env.local. Keep these files, your workspace, and your backups private; never commit credentials.
The installers validate the key against the API. Read-only calls such as socials list are a good first check. Creating, updating, scheduling, or deleting resources changes real FeedHive data. Give your agent clear approval boundaries for destructive or hard-to-undo changes.
- FeedHive public API: endpoint payloads, validation rules, and response shapes.
- CLI reference: commands and examples for
@feedhive/cli. - Bundled script guide: lower-level Node.js scripts used by the skills.
- OpenClaw skill and Claude Code skill: agent instructions and available operations.
- Package-specific READMEs: what each npm package installs or runs.
- Hosted MCP server: connection setup and tool reference for
https://mcp.feedhive.com.
The repository contains the source for the skills, scripts, CLI, setup packages, and tests. The FeedHive server, including the hosted MCP server, lives outside this repository. For a local source build, run npm ci, npm test, npm run typecheck, and npm run build. npm releases are manually triggered by maintainers; pushing a commit does not publish a package. To release, run the Publish selected npm package (manual) workflow from main, select one package, and enter a new npm version. Run it separately for each package; no Git tag is required.
Licensed under the MIT License. Please report suspected security issues privately to FeedHive rather than posting credentials in an issue.