Agentic Use (MCP, CLI, Skill)
Choose the right way to connect AI agents and automation tools to TalentSourcer AI.
Here is the simple cheat sheet:
| Option | Who it is for | When to pick it | How you log in |
|---|---|---|---|
| Remote MCP (recommended for most) | Recruiters who use ChatGPT, Claude, Cursor, or another MCP-compatible AI assistant | You want to use TalentSourcer AI directly inside your favorite AI chat. Fastest and easiest setup. | Browser login with OAuth |
| CLI | Recruiters, ops teams, or technical users who use the terminal or scripts | You want to run TalentSourcer AI from your terminal, automate recurring workflows, or integrate it into internal tools. | Personal access token |
| Skill | Users working with AI coding agents like Cursor, Codex, or OpenCode | You want your local AI agent to understand how to use the TalentSourcer CLI safely and consistently. | Uses the CLI after you authenticate once |
Quick decision guide
Start with Remote MCP if...
This is what most recruiters should use.
Use Remote MCP if you open ChatGPT, Claude, Cursor, or another AI assistant regularly and want to ask it to work with TalentSourcer AI directly.
Example requests:
- “Show me the strongest candidates for this project.”
- “Create a shortlist from the top completed candidate checks.”
- “Summarize the talent pool insights for this briefing.”
- “Draft an outreach campaign for this shortlist.”
Use the CLI if...
- You are comfortable in the terminal.
- You want to run the same workflow repeatedly.
- You are building small scripts or internal automations.
- You need structured JSON output for another tool.
Use the Skill if...
Use the Skill if you spend time inside Cursor, Codex, OpenCode, or a similar coding-agent environment and want that agent to know how to use TalentSourcer AI correctly.
The Skill is especially useful when the agent needs to create local JSON payload files, run CLI commands, and chain TalentSourcer actions together.
Bottom line
Most recruiters should start with Remote MCP.
Only use the CLI or Skill if you specifically need terminal automation or you work heavily inside an AI coding-agent environment.
What agents can do
Agentic access can help with recruiting workflow operations such as:
- listing, creating, and updating projects
- reading and updating briefings and requirements
- importing candidates into searches
- reviewing candidate checks
- creating and updating shortlists
- creating draft campaigns and editing campaign flows
- enrolling shortlists into draft campaign workflows
- reading campaign analytics
- reviewing, snoozing, skipping, or marking outreach tasks as sent
Campaign and outreach actions are intentionally safe by default. Agents can create drafts and update workflow records, but they do not automatically launch campaigns, send email, or send LinkedIn messages.
Recommended scopes
Choose the smallest scopes your workflow needs.
| Workflow | Scopes |
|---|---|
| Read projects and briefings | projects:read, briefings:read |
| Create or update projects and briefings | projects:write, briefings:write |
| Import or review candidates | candidates:read, candidates:write, ats:read, ats:write |
| Manage shortlists | shortlists:read, shortlists:write |
| Create and edit draft campaigns | campaigns:read, campaigns:write |
| Review manual outreach tasks | outreach_tasks:read, outreach_tasks:write |
| Keep OAuth MCP access connected | offline_access |
For broad recruiting-agent workflows, use:
projects:read projects:write briefings:read briefings:write candidates:read candidates:write shortlists:read shortlists:write campaigns:read campaigns:write outreach_tasks:read outreach_tasks:write inbox:read ats:read ats:write offline_accessSecurity guidance
- Use OAuth for MCP clients whenever possible.
- Use personal access tokens only for CLI and technical automation.
- Do not paste access tokens into AI chat windows.
- Revoke tokens you no longer use from TalentSourcer AI settings.
- Create separate tokens for separate agents or machines.
- Prefer read-only scopes when the agent only needs to inspect data.