Roadmap · Agent-to-agent

From copilot to the backlink agent other agents hire.

Today you approve every send. The road ahead trades clicks for constraints: you set the outcome, the standards, and how far Seowly may go on its own — it does the work, and proves every link it earns.

V1 · Shipping July 26, 2026A human approves every sendSent from your own inbox

The path

The shift

Delegate an outcome, not a sequence of clicks.

Every link tool on the market asks you to operate it. The end state asks you for something else entirely — a goal, a standard, a budget, and the authority to act inside them.

Today

Seowly does the backlink work; you approve.

You open the app, review what it found, edit the draft, and press send. Seowly carries the effort. You carry every decision.

Where this goes

Tell Seowly the backlinks you want, the standards it must follow, and the authority it has. It handles the rest — and proves every result.

Your agent hands ours a mandate. Seowly plans, executes, escalates the calls that need you, and returns evidence for everything it claims.

Authority in rungs, not one autopilot switch

A single on/off toggle can't carry trust. Authority is granted a rung at a time — and Seowly stops the moment an action falls outside it.

Observe
Discover, analyse, and recommend. Nothing leaves the building.
Live in V1
Prepare
Qualify opportunities and prepare outreach and execution plans.
Live in V1
Execute free actions
Send permitted outreach automatically, inside your limits.
Granted later
Negotiate
Handle replies and counteroffers within defined constraints.
Granted later
Spend
Commit funds below explicit per-placement and total caps.
Granted later
Escalate
Pause and ask you the moment a decision falls outside the mandate.
Granted later

The path

Six steps to a link-building agent.

V1 has a launch date. Everything after it has an order, not a calendar — each step ships when the one before it has earned enough trust.

V1Shipping July 26, 2026
Supervised copilot
The whole loop, with you in the approval seat.
- Discovery from your competitors' backlinks, plus a mid-cycle keyword and SERP sweep.
- Contacts found by scraping the prospect's own site — no paid email finder.
- Drafts written per strategy: guest post, directory, mention reclamation, link exchange.
- Nothing leaves your inbox until you approve it — and the same message is never sent twice.
- Replies detected, follow-ups auto-cancelled, backlinks reconciled against your profile.
- Versioned discovery and copywriting modules; every cost-incurring action metered and capped.
V1.2Next
Connect any inbox
"Your own inbox" stops meaning "your Gmail".
- Gmail, Outlook and Microsoft 365, or any IMAP host — one connect flow for all of them.
- Send-only access. Seowly never asks for permission to read your mailbox.
- Replies route through a Seowly address, so detection works identically on every provider.
- A dedicated outreach inbox is recommended in onboarding — never enforced.
V2Planned
Trust and controlled autonomy
A policy engine, not an autopilot toggle.
- Confidence thresholds decide what sends itself and what still waits for you.
- Replies parsed into structured offers a policy can accept, decline, counter, or escalate.
- An exception inbox replaces the approval queue — you only see the judgment calls.
- A complete decision audit trail: what Seowly did, and why it did it.
- Webhooks, idempotent asynchronous jobs, and a dry-run mode.
- Results measured per strategy and module version. Optionally bring your own data and AI keys.
V2.5Planned
Agent-accessible
Everything a human can do in the dashboard, an agent can do through the API.
- A public REST API, described end to end by OpenAPI.
- Scoped agent credentials, with explicit permissions and side effects.
- llms.txt, Markdown documentation, and written guides for common mandates.
- An installable Seowly skill and an MCP server, published to agent registries.
- A real sandbox: rehearse a whole integration without sending an email or spending a cent.
- A structured feedback endpoint, so agents can tell us exactly where we fell short.
V3Where this is going
Delegated outcomes
The Backlink Mandate becomes the product.
- One object carries the objective, the quality policy, the outreach policy, and the budget.
- Graduated authority from observe through spend, enforced by the policy engine.
- Autonomous reply handling and negotiation, strictly inside your constraints.
- A scoped wallet: mandate budget, per-placement cap, reserved funds, immutable ledger.
- An evidence bundle for every link claimed. No bundle, no result.
- The dashboard becomes a command centre: what it's doing, what it may do, what it spent.
V3+Exploring
The backlink opportunity network
Seowly as the trust layer between brand mandates and verified publishers.
- Verified publisher identities and machine-readable editorial availability.
- Reputation that runs both ways — for buyers and for publishers.
- Automated matching and negotiation, agent to agent.
- Protected payment, delivery disputes, and remediation.
- Cross-network backlink verification and survival monitoring.

V3 · The core object

The Backlink Mandate

Not a campaign. Not a workflow. A mandate is what you want, the standard it has to meet, and how far Seowly may go on its own to get there.

What you say

Build 10 high-quality, editorially legitimate backlinks for seowly.com. Spend no more than $2,000. Avoid these topics and domains. Ask me only when a decision falls outside policy.

The same instruction, machine-readable — the object every API call, MCP tool, notification, and dashboard surface revolves around.

Objective, quality, outreach, and budget are separate policies on purpose. Loosen one without touching the others.

What your agent sends

POST /mandates

{
  "objective": {
    "backlinks": 10
  },
  "qualityPolicy": {
    "minimumTopicalRelevance": 0.8,
    "minimumAuthority": 40,
    "minimumOrganicTraffic": 1000,
    "allowedStrategies": [
      "guest_post",
      "mention_reclamation",
      "editorial_inclusion"
    ],
    "blockedDomains": [],
    "blockedTopics": []
  },
  "outreachPolicy": {
    "dailySendLimit": 8,
    "autoApproveConfidenceAbove": 0.92,
    "maxFollowUps": 2
  },
  "budgetPolicy": {
    "maximumTotal": 2000,
    "maximumPerPlacement": 250,
    "requireApprovalAbove": 150
  }
}

V2.5 · Agent-ready

Seven layers between a website and a service an agent can hire.

Being agent-ready is more than shipping an MCP server. An agent has to find Seowly, understand it, delegate to it, operate it, pay it, trust it, and measure it.

Discoverable

A public API, an MCP server, installable skills, and registry listings.

Legible

llms.txt, OpenAPI, Markdown docs, explicit semantics, real examples.

Delegable

A machine-readable mandate: outcome, constraints, authority, budget.

Operable

Async jobs, webhooks, stable lifecycle states, retries, idempotency.

Transactable

Subscriptions, usage credits, placement payments, a scoped wallet.

Trustworthy

A policy engine, evidence, an audit trail, dry runs, escalation rules.

Measurable

Task completion, backlink survival, cost, and return — reported back.

Three ways in

REST API + OpenAPI

The durable foundation. Every action in the human dashboard runs through the same application API — no UI-only business logic hiding anywhere.

MCP server

A short list of outcome-shaped tools: estimate a mandate, authorize it, check status, review an exception, fetch evidence. Not a database with a wrapper on it.

Installable skill

A skill file that teaches an agent when to use Seowly, what a good mandate looks like, how we define link quality, and which mandates are unsafe to run.

Rehearse before you commit

Before an agent sends an email or spends a cent, it can run the whole mandate as a dry run against a demo brand and synthetic prospects.

No real email. No paid providers. No wallet movement. No publisher ever contacted.

Feasibility: Medium
Opportunities found: ~128
Outreach messages: ~42
Backlinks expected: 4–11

The receipt

"Task completed" is not a result.

A backlink claim is only worth what backs it up. Every link Seowly reports carries an evidence bundle — and keeps getting re-checked long after it's won.

The operating loop

01Agent delegates a mandate
02Seowly proposes a plan
03Policy authorizes each action
04Discovery and outreach run
05Negotiate, or escalate to you
06Backlink acquired
07Verified and recorded as evidence
08Agent evaluates the outcome

Paid placements are always labelled as paid. Evidence preserves a link's real attributes and disclosure status — we will never imply a sponsored link carries the weight of an earned one.

Constant

Five things the roadmap never changes.

It sends from your own inbox.
No rented cold-email farm, no warmed-up throwaway domains. Your sender reputation stays yours.
A human can always be the last word.
Autonomy is something you grant, one rung at a time — and take back in a single click.
Free links first.
Competitor gaps, directories, communities, reclaimed mentions. Paid placement is an option, never the default.
Evidence over claims.
Nothing counts as a link until it's verified — and it keeps being re-verified after that.
The same message is never sent twice.
Not on a retry, not across brands, not when two agents ask for the same thing.

V1 ships July 26, 2026. Start there.

Seowly is the backlink agent other agents hire to acquire high-quality links within a defined budget and quality policy. It begins as the copilot you approve.

Start free

3-day trial · no charge today · cancel anytime

From copilot to the backlink agent other agents hire.

Delegate an outcome, not a sequence of clicks.

Authority in rungs, not one autopilot switch

Six steps to a link-building agent.

Supervised copilot

Connect any inbox

Trust and controlled autonomy

Agent-accessible

Delegated outcomes

The backlink opportunity network