One loop, five pieces.
Harbor remembers how each brand sounds, plans the week from what worked, drafts it for every platform, publishes at the right hours and learns from every post. Each piece feeds the next.
Each brand keeps its own memory.
Voice, audiences, products, knowledge and style rules, per brand. Start by picking sample posts that sound right, then let results refine it: the voice refresh proposes rules with the posts behind them, and you approve each change.
- Learned from posts you pick, paste or import
- Every rule links to its evidence
- Preview a draft with and without memory
Agents plan the week. You approve it.
The weekly plan reads last week’s digest, finds open best-time slots, suggests ideas, drafts them and checks each draft against the voice. What it makes waits in Approvals with the reason for every slot.
- Five built-in agents, every run logged
- Automations on a schedule or on post events
- Claude can work in Harbor over MCP
- …Read last weekDigest: 9 posts, median 2.9K views at 24h (+22%).3ms
- Find open slots6 open slots over the next 7 days on X, LinkedIn, Threads.11ms
- Suggest ideas3 ideas guided by 4 recommendations.1.2s
- Draft every idea6 drafts, one per platform, in the brand’s voice.3.4s
- Voice check6 of 6 pass: banned words, length, hashtags, media.8ms
- Queue for approval6 posts waiting for you. Nothing published.4ms
One idea, drafted for every platform.
Each platform gets a draft in its native shape and limits, with a live preview. Ask for rewrites, keep versions, and lay out a campaign on a canvas where connected cards become the agent’s context.
- Ideas cite the signals behind them
- Threads, carousels and per-part counts
- A board or table per campaign
Out at the hours that work.
Direct to X, LinkedIn, Instagram and Threads through each platform’s API, checked against its rules and cropped from the media library. The calendar is shaded by this brand’s best times; drag a post to move it.
- Sandbox mode for free, labeled practice runs
- Focal-point crops for every platform
- Never retries a post it can’t confirm
What worked, and why.
Every post is measured at 1h, 6h, 24h, 3 days and 7 days against the brand’s own median at the same age. The weekly digest says what worked, what missed and what to do next, and feeds the next plan.
- Outperformed, met, underperformed or inconclusive
- Best times as a 7 × 24 map per brand
- Reply themes become signals for new ideas
Every result shapes the next plan.
Most tools stop once a post is out. In Harbor, what a post did flows back into the brand’s memory and the next week’s plan. Pick a stage to see what it reads and what it writes.
Every brand’s voice, audiences and offers, learned from what actually performs.
Explore Brand Memory →- Posts you pick, paste or import
- Voice-refresh changes you approve
- Anything you save from a post or reply
- Voice, audiences, products, knowledge and style rules
- Evidence: the posts behind each rule
Plan next week in an afternoon.
Self-host Harbor free, or get early access to Harbor Cloud. Either way, nothing posts until you say so.