Agents

Agents that do the strategist's homework

Each one chains what Harbor already knows: Brand Memory, the digest, best times and every post's growth curve. Nothing posts on its own. What an agent drafts waits in Approvals until you say yes.

Connect Claude (MCP)Approvals

Simulated Agents run on the simulated model: real steps and real data, no API spend.

Plan

Weekly plan

Plan next week from what worked last week.

  1. Read last week
  2. Find open slots
  3. Suggest ideas
  4. Draft every idea
  5. Voice check
  6. Queue for approval
ProducesDigestIdeasDraftsApprovals
RunsRun by hand · On a schedule
Not run yetAutomateRun
Repurpose

Repurpose a winner

Turn a post that outperformed into posts for your other platforms.

  1. Read the winner
  2. Turn it into an idea
  3. Draft for other platforms
  4. Voice check
  5. Find open slots
  6. Queue for approval
ProducesIdeasDraftsApprovals
RunsRun by hand · When a post outperforms
Not run yetAutomateRun
Review

Voice check

Check a draft against the brand voice and platform rules.

  1. Check the draft
ProducesReport
RunsRun by hand
Not run yetRun
Insights

Weekly digest

What worked, what missed, and what to do next week.

  1. Write the digest
ProducesDigest
RunsRun by hand · On a schedule
Not run yetAutomateRun
Review

Voice refresh

Learn the voice from what actually worked.

  1. Refresh verdicts
  2. Compare winners and losers
  3. Check the current rules
  4. Draft the changes
  5. Send for approval
ProducesReportApprovals
RunsRun by hand · On a schedule
Not run yetAutomateRun
Bring your own agent

Claude, or any MCP client, can read Brand Memory, suggest ideas, draft and propose posts through Harbor's MCP server. The same rule applies: proposals land in Approvals.

Create an API key and connect →