Skip to content
3

Agent Access

Connect the tools your team already uses. first(20) is the hiring layer underneath your agents. You decide what they can read, write, draft, and approve. Every action is logged.

first(20) is not trying to become your new interface.

Give Claude, ChatGPT, Cursor, or your internal agent permission to operate recruiting safely. Humans see, review, and approve. Agents operate.

Connected agents

Claude
ClaudeActive
Connected via
MCP
Permissions
6
Actions this week
38
Last used
20m ago
OpenAI
ChatGPTActive
Connected via
API
Permissions
4
Actions this week
21
Last used
3h ago
Cursor
CursorIdle
Connected via
MCP
Permissions
2
Actions this week
4
Last used
2d ago
Recruiting AgentActive
Connected via
API
Permissions
5
Actions this week
52
Last used
5m ago
Internal Ops AgentActive
Connected via
API
Permissions
4
Actions this week
9
Last used
9h ago

MCP connection

first(20) MCP Server

https://mcp.first20.io/sse

14 tools exposed

Connected

API keys

  • Production: Recruiting Agent

    f20_live_••••••••8t2k

    Created Apr 2, 2026 · last used 5m ago

  • Read-only: Ops dashboards

    f20_live_••••••••q19v

    Created May 10, 2026 · last used 9h ago

  • Staging: experiments

    f20_test_••••••••0c4d

    Created Jun 1, 2026 · never used

What agents can operate

Read jobs and candidates · update candidate stages · draft emails · request feedback · summarize candidates · generate pipeline reports · create job posts · update scorecards · read funnel stats · write audit log entries.

Agent permissions

Permissions are the product. Even one agent operating recruiting means access must be explicit, scoped, and revocable.

Read

View roles, requirements, and pipeline health.

jobs:read
Write

Create job posts, edit requirements, and publish roles.

jobs:write
Read

View candidate profiles, scorecards, and stage history.

candidates:read
Write

Move candidate stages and update candidate records.

candidates:write
Write

Chase interviewers and hiring managers for missing feedback.

feedback:request
Write

Draft candidate follow-ups and rejections for approval.

emails:draft
Read

Produce hiring updates and funnel reports.

reports:generate
Read

Read the full history of human and agent actions.

auditLogs:read
Write

Create and revoke API keys and MCP connections.

apiKeys:manage
Write

Approve or reject pending agent actions.

approvals:approve

Audit logs

Full activity
  1. Recruiting AgentDrafted follow-up emailsPending approval
    Email: 4 candidates · Drafted, pending approval
  2. ClaudeSummarized candidatesAuto-approved
    JobPost: Founding Engineer · 12 summaries generated
  3. Priya NairApproved stage changeApproved
    Candidate: Maya Chen · Interview → Final
  4. Recruiting AgentMoved candidate stageApproved
    Candidate: Maya Chen · Interview → Final
  5. Jordan BlakeCreated API keyApproved
    ApiKey: Staging: experiments · Created with read scopes
  6. Internal Ops AgentGenerated hiring reportAuto-approved
    HiringReport: Weekly update · Report generated