Topo
Run your Topo outbound workspace from
Claude,
ChatGPT and Codex, Cursor,
and other Agent Plugins clients. The plugin
connects the Topo MCP server and loads the skills
sales teams run every day — when to call stop_sequence vs unenroll, which
id to pass, and the rest of the product rules the tools alone do not carry.
A Topo account is required. Connecting MCP without this plugin exposes tools only; it does not load the workflows.
What you can do
- Triage inbound replies and answer them in-thread.
- Record a booked meeting, a win, a loss, or an unsubscribe so sending stops.
- Review the leads waiting on a sequence and approve or refuse them.
- Clear today's tasks, snooze what can wait, pause or resume outreach.
- Import people from a list or CSV and start a sequence.
- Create or clone a sequence template and edit its copy.
- Source new leads from criteria with a preview before credits are spent.
- Suppress or release emails and domains.
- Brief an account from its contacts, enrollments, and threads.
- Diagnose a quiet sequence and apply the operational fixes MCP allows.
Reads are free. Anything that sends, stops, blocks, or spends credits asks for your confirmation first.
Install
Claude Code
/plugin marketplace add topo-io/topo-plugin
/plugin install topo@topo-plugins
Local test from a checkout:
claude --plugin-dir .
claude plugin validate . --strict
ChatGPT and Codex
Install Topo from the plugin directory, or point a local marketplace at a
checkout of this repository (plugin.json at the root is the portable
manifest; .codex-plugin/plugin.json is the compatibility fallback).
Other Agent Plugins clients
Load the directory: plugin.json, mcp.json, and skills/ follow Agent
Plugins 1.0.0.
Sign in
The first tool call opens a Topo sign-in (OAuth). The plugin only sees the workspace your Topo user belongs to. Remove the plugin from your host to revoke access; tokens are deleted with it.
Skills
| Skill | Job |
|---|---|
getting-started-with-topo | Confirm the connection, take stock, and route to the right skill |
triaging-inbound-replies | Work inbound email/LinkedIn replies and apply the matching outcome |
recording-outreach-outcomes | Apply a known win/loss/stop without corrupting the funnel |
reviewing-waiting-leads | Qualify waiting leads; approve or refuse without a 3-month block |
clearing-todays-tasks | Clear the personal task queue with the type-specific payload |
importing-and-starting-outreach | Put named people on a list and enroll them in an existing sequence |
creating-a-sequence | Author a new sequence, clone one, or replace its steps and copy |
creating-a-lead-search | Source new leads from criteria, preview or run, and import into a list |
suppressing-or-releasing-targets | Block or lift emails/domains and stop anything already sending |
briefing-an-account | Assemble who we know, who is in sequence, and what they said |
diagnosing-sequence-health | Explain why a sequence is quiet and apply allowed ops fixes |
holding-or-resuming-outreach | Pause for OOO / bad timing without marking people lost |
ids-and-outcomes | Which id goes to which tool, and which write records each outcome; every other skill reads it first |
Layout
| File | Reader |
|---|---|
plugin.json + mcp.json + skills/ + assets/ | Agent Plugins 1.0.0, ChatGPT, Codex |
.claude-plugin/plugin.json + .mcp.json | Claude Code |
.codex-plugin/plugin.json | ChatGPT / Codex compatibility fallback |
MCP server: https://mcp.topo.io/mcp.
Support, privacy, terms
- Support: https://support.topo.io
- Documentation: https://docs.topo.io
- Plugin issues: https://github.com/topo-io/topo-plugin/issues
- Privacy policy: https://www.topo.io/consents/privacy-policy
- Terms of service: https://www.topo.io/consents/terms-of-service
Source of truth
Authored in the private topo-io/agent
monorepo (packages/topo-plugin/) and mirrored here so the plugin directories
can review a public GitHub URL. See CHANGELOG.md for releases.