
whappy
ai.whappy/whappy
Build and rehearse WhatsApp and Instagram AI-setter campaigns, and read leads, chats and results.
Documentation
Whappy plugins for Claude Code and Codex
Whappy runs AI setters on WhatsApp and Instagram: they answer your leads, ask your
qualifying questions and book the appointment. This plugin connects Claude Code or
Codex to your Whappy account through the Whappy connector
(https://mcp.whappy.ai/mcp) and adds three workflows:
| Skill | What it does |
|---|---|
build-a-campaign | From a website or a business description to a built, checked and rehearsed campaign. You launch it in the dashboard. |
daily-inbox-review | Unread chats, chats waiting on a human, chats in error, and what to do with each. |
weekly-report | Leads, conversations, appointments, the funnel, Coach suggestions and next steps for the week. |
Two commands run the last two directly: /whappy:inbox and /whappy:report.
The plugin holds no secrets and no hooks. It only declares the connector and the instructions above; you sign in to Whappy in your browser the first time a tool runs.
Install in Claude Code
In a Claude Code session:
/plugin marketplace add whappyAI/whappy-ai-plugins
/plugin install whappy@whappy-ai-plugins
From a shell, the same steps are
claude plugin marketplace add whappyAI/whappy-ai-plugins and
claude plugin install whappy@whappy-ai-plugins.
Then run /mcp, pick plugin:whappy:whappy and sign in. The skills appear as
/whappy:build-a-campaign, /whappy:daily-inbox-review and /whappy:weekly-report,
and Claude also uses them on its own when your request matches.
Install in Codex
codex plugin marketplace add whappyAI/whappy-ai-plugins
Then open the plugin directory in Codex (/plugins), choose the Whappy
marketplace and install Whappy. Codex signs you in to Whappy when you install it.
Without the plugin, the connector alone is
codex mcp add whappy --url https://mcp.whappy.ai/mcp followed by
codex mcp login whappy.
Other assistants
Claude on the web, desktop and mobile, ChatGPT and other MCP clients can add the
connector by URL: https://mcp.whappy.ai/mcp. See
Claude and other MCP clients. Gemini CLI
has its own extension: whappyAI/gemini-cli-extension.
Permissions
When you connect, Whappy shows a consent page with the client that is asking, the account you are signed in with and where you will be sent back. You choose Allow or Deny; there is no access level to pick.
A connected assistant can:
- build a campaign from a website or a description, and create, rename, duplicate or delete campaigns;
- read, generate and edit the funnel, the knowledge and the qualifying questions;
- draft WhatsApp templates, translate them, submit them to Meta and send a test template to a phone number you give it;
- change a campaign's working hours, time zone and start delay;
- rehearse a campaign with synthetic leads, who are never messaged;
- read your leads, conversations, results, Coach reports, calendar settings and plan.
It cannot start or stop a campaign, message your leads, add leads, change billing or your account, or connect WhatsApp or Instagram (it gives you a link to finish in the browser). The tools that delete a campaign, a knowledge document, a question or a funnel step are marked destructive, so assistants that honour the mark, such as Claude and Codex, ask you before calling them. AI generation is limited to 30 requests per hour per account, shared with the assistant in the dashboard.
Some tools the skills mention are marked "when available": they are planned and not yet offered by the connector. The skills use them only once the connector lists them.
To disconnect, remove the connector in your client or uninstall the plugin; the client tells Whappy to revoke the access. There is not yet a dashboard page listing your connections: if you lost access to the client, contact support and we revoke it.
Support and privacy
- Support: docs.whappy.ai/support/contact-support or team@whappy.ai
- Documentation: docs.whappy.ai/integrations/mcp
- Privacy policy: whappy.ai/legal/privacy
- Terms: whappy.ai/legal/terms
Data you read through the connector goes to the AI vendor of the assistant you connected, under your own agreement with that vendor.
MCP Registry
server.json lists the connector in the official MCP Registry as ai.whappy/whappy
(remote only, Streamable HTTP). The namespace ai.whappy is proven by a DNS TXT
record on the apex of whappy.ai:
whappy.ai. IN TXT "v=MCPv1; k=ed25519; p=Y3WT2o9ZEA8vhGpj1l+CfWTP7Xxyel7aSaS4+MZBM3A="
Add it at the apex (whappy.ai, host @), not under a subdomain. It can sit next to
other TXT records such as SPF. Check it with dig TXT whappy.ai +short.
The matching Ed25519 private key is not in this repository. Store it, as the hex
string that mcp-publisher login dns --private-key takes, in the GitHub Actions
secret MCP_PRIVATE_KEY on an environment named mcp-registry-publish, limited to
v* tags and with a required reviewer.
.github/workflows/publish-mcp-registry.yml then publishes on every v* tag whose
version matches server.json.
Releasing
- Change the version in
.claude-plugin/plugin.json,.codex-plugin/plugin.jsonandserver.jsontogether (a test checks that they agree). - Merge to
mainwith CI green. - Tag
vX.Y.Zand push the tag. The registry workflow publishesserver.json. Claude Code and Codex users get the new version when they update the marketplace.
Development
claude plugin validate . --strict # plugin and marketplace (Claude Code)
python3 -m unittest discover -s tests -v # URL, manifests, skill tool names, no secrets
npx --yes markdownlint-cli2 # Markdown lint
CI (.github/workflows/ci.yml) runs these, plus a JSON schema check of the Codex
manifests (schemas/) and mcp-publisher validate server.json.
tests/connector_tools.json is the list of tools the connector exposes, copied from
the Whappy backend's connector profile. A skill may name only a listed tool, and a
planned tool only on a line that says "when available". Update the list when the
connector changes.
Layout
.claude-plugin/plugin.json Claude Code plugin manifest
.claude-plugin/marketplace.json Claude Code marketplace (this repo)
.codex-plugin/plugin.json Codex plugin manifest
.agents/plugins/marketplace.json Codex marketplace (this repo)
.mcp.json the Whappy connector, shared by both
server.json MCP Registry entry
skills/ the three skills
commands/ /whappy:inbox and /whappy:report
schemas/ JSON schemas for the Codex manifests
tests/ contract tests
Sources
File formats checked against the vendors' documentation on 2026-11-22:
- Claude Code plugin manifest: https://code.claude.com/docs/en/plugins-reference
- Claude Code components (skills, commands,
.mcp.json): https://code.claude.com/docs/en/plugins/components.md - Claude Code marketplaces: https://code.claude.com/docs/en/plugin-marketplaces and https://code.claude.com/docs/en/plugins/marketplace-reference.md
- Codex plugins and marketplaces: https://developers.openai.com/plugins/build/plugins.md
- Codex manifest parser: https://github.com/openai/codex/blob/cf3ff1530d821ec2d26452da4971d8e3e7d18d25/codex-rs/core-plugins/src/manifest.rs and
marketplace.rsin the same directory - Codex compatibility examples: https://github.com/openai/plugins/tree/main/plugins/notion
- MCP Registry, remote servers: https://modelcontextprotocol.io/registry/remote-servers.md
- MCP Registry, DNS authentication: https://modelcontextprotocol.io/registry/authentication.md
- MCP Registry, GitHub Actions: https://modelcontextprotocol.io/registry/github-actions.md
server.jsonschema: https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json
License
MIT, see LICENSE.
No installation packages available.