Standing - Turn Projects Into Predictable Monthly Revenue - Vertical Agent Spec

One-line definition

An agent that converts a freelancer's existing project-based client relationships into scoped, auto-renewing retainer agreements by drafting, sending, and tracking the conversion proposal cycle.

The workflow it owns end-to-end

Ingests the freelancer's current client list, project history, and hourly or per-project billing records to identify which clients have recurring needs and predictable work volume
Calculates a retainer scope recommendation per client based on trailing 90-day hours, deliverable types, and typical month-over-month variance
Drafts a retainer proposal in the freelancer's voice, including scope definition, rollover policy, pricing rationale, and a simple "yes, let's do it" response path
Sends the proposal via email or a tracked link, monitors open and click events, and generates a follow-up nudge sequence if no response within a set window
On client acceptance, generates a lightweight service agreement and triggers the first invoice through the connected payment processor

What it knows that a generic LLM doesn't

The specific vocabulary of retainer structures that don't frighten project clients: "reserved hours," "monthly scope," and "priority access" versus "retainer," which carries legal connotations many freelancers and small clients find intimidating
The math behind rollover policies and how to frame caps (for example, unused hours roll over once, cap at 1.5x monthly) in ways that protect the freelancer's time without making the client feel penalized
Which client signals actually predict retainer readiness: three or more consecutive months of work, inbound "can you just handle X ongoing" language, and clients who contact the freelancer outside normal project windows
Common objections at the proposal stage ("I don't always need you every month" and "I don't want to pay for hours I don't use") and the specific counter-framings that close them without overselling
Scope creep patterns by freelancer category (design, development, copywriting, bookkeeping) and how to define deliverable caps that hold up in practice
The difference between a retainer proposal for a $500/month client versus a $5,000/month client in terms of formality, contract depth, and response timeline expectations

What it explicitly declines

Negotiating on the freelancer's behalf in real time during a client call or email thread; the agent prepares materials, it does not participate in the live conversation
Providing legal review or certifying that a service agreement is enforceable in any jurisdiction; it generates a working document, not a lawyer-reviewed contract
Making pricing decisions without freelancer input; it surfaces a recommendation with reasoning but requires explicit approval before sending any dollar figure to a client
Managing the ongoing retainer relationship after it is signed; scope disputes, late payments beyond a first reminder, and client dissatisfaction are handed off to the human immediately

Tools and integrations required

Stripe or Wave for invoice generation, recurring billing setup, and payment status monitoring
Gmail or Outlook via OAuth for proposal delivery and reply detection, with open and click tracking through a service like Postmark or SendGrid
Bonsai, HoneyBook, or a generic DocuSign integration for agreement delivery and e-signature collection
A lightweight CRM layer, either a connected tool like HubSpot or an internal Airtable base, to track proposal status per client across the pipeline
Calendly or Cal.com for optional "let's talk through this" booking links embedded in proposals when the agent detects the client profile suggests a conversation is needed before signing
Google Docs or Notion for storing and versioning the freelancer's master proposal templates and scope language so the agent personalizes without rewriting from scratch each time

Trust escalation: when it pings a human

When a client replies to a proposal with any language that reads as a counter-offer, complaint, or legal question, the agent flags the thread and stops the follow-up sequence until the freelancer reviews and responds
When the proposed retainer price exceeds a freelancer-set threshold (for example, $3,000/month), the agent drafts but does not send without explicit approval, because a misjudged proposal at that level damages the relationship more than the missed conversion hurts
When a client has not opened the proposal after two follow-ups and the agent would otherwise send a third, it pauses and surfaces the client for a manual decision, since persistent automated follow-up with unresponsive clients risks the underlying relationship
When the calculated scope recommendation would require the freelancer to commit more than 80 percent of their available hours to a single client, the agent flags the concentration risk before proceeding

Pricing model

The honest pricing model here is per-converted-retainer, not per seat. Charge $49 per retainer agreement that results in a signed contract and a first successful payment, with a $19/month platform fee for active users who want the follow-up sequencing, template storage, and CRM sync running continuously. At an average freelancer retainer of $2,000/month, a $49 success fee is a straightforward value exchange. The $19/month base keeps the agent viable during slow months without making users feel like they are paying for nothing. The risk is that the success fee model requires reliable payment attribution, and if a freelancer closes a deal over the phone after the agent sent the proposal, tracking that conversion honestly is technically and contractually messy. This pricing is directionally correct but needs a clear attribution window, something like: a signed contract within 30 days of agent-sent proposal counts as a conversion.

Differentiation from a generic LLM wrapper

The reason a freelancer would use this instead of pasting their client list into Claude is not the writing quality; Claude writes a perfectly good retainer proposal. The differentiation is state management and workflow continuity across weeks. A generic LLM produces a document and then forgets. This agent tracks which clients received proposals, which ones opened them, which ones replied, which ones went cold, and when to follow up with which message given where each client sits in the cycle. It also holds the freelancer's pricing history, scope templates, and client-specific context so the fifth proposal it writes is better calibrated than the first. That said, the activation wall identified in the known risks is real and is not solved by better writing. If the hard part is the sales conversation and the agent does not participate in that conversation, a meaningful percentage of users will blame the tool when deals do not close, regardless of proposal quality. The agent shape works for proposal preparation and pipeline tracking. It does not work as a closing tool, and the product should not imply that it does.