Describe the job. Your agent does it on schedule, inside your rules
Write a brief like you would for a new teammate, then choose when it runs, which apps it may use and what needs your approval. Your agent researches, qualifies, writes, updates and reports, and shows you exactly what it did, step by step, with the credits each step used.
- 7 days free
- No card required
- Nothing sends without your yes
Your job.Your rules.Its shift.
Every team has the job it does by hand each week: the report, the re-check, the list nobody updates. Describe it once. Your agent picks it up on schedule or the moment something happens, uses the apps you already pay for, and asks before anything leaves your hands.
What the Custom agent does, step by step
The same steps your best teammate would take, written down once and run every time.
-
01
Write the brief
What to look at, what to do, when to stop and what to report. Start from a template, from scratch, or let Help me write this draft it from what your workspace knows.
-
02
Pick when it runs
On demand, on a schedule, or when something happens: people added to a list, a labelled reply, someone turning high intent, an import finishing.
-
03
Give it skills and apps
Attach playbooks as skills, eight built in or up to fifty of your own, and tick the connected apps it may use.
-
04
Set the guardrails
Choose what needs approval, which actions it may repeat on its own, and the most steps and credits one run may use.
A run page that says exactly what happened
No logs to decode. Every run reads like a note from a colleague, with the result, the credits and anything waiting for you.
- What it did, in plain wordsEach step listed with the credits it used.
- What is waiting for youApprove or decline right on the run page.
- The resultFormatted answers, tables and diagrams to copy or download.
On your schedule, or the moment something happens
Every trigger is available: on demand, on a schedule, and on events in your workspace.
- When you run itDefault
- On a schedule
- When people are added to a list
- When a labelled reply comes in
- When someone turns high intent
- When an import finishes
Up to two runs are active at a time in a workspace; others wait their turn. A run pauses at its step or credit limit until you continue it.
The Custom agent uses the tools you already pay for
Tick the connected apps it may use. Reads run straight away; anything that changes data in another app waits for your yes.
Custom agent asks before it acts
You choose how much it does alone. Reading and research always run straight away; anything that sends, spends or changes another app waits as a card you approve, unless you allowed it.
- Review each stepIt asks before every step, unless you allowed that action to run on its own.
- Review changesIt reads and researches freely, then brings every change back to you.
- Run routine workSaving people, tagging and preparing drafts run on their own. Sending, paid actions and other apps still wait.
Put the Custom agent to work
Real jobs, with the request, the steps and the result. Open one and copy it into your workspace.
New people on a listFill and verify missing work emails with Hunter.io before you launchBefore an email campaign goes out, an agent looks up the people on your list who have no email in Hunter.io, keeps only verified addresses and asks before saving them.Custom agent
On demandAI account research: a sourced brief on every target accountAn agent runs deep research on each target account with Parallel and Exa, and hands you a one-page brief with priorities, news, people and a reason to reach out, every claim with its source.Custom agent
New people on a listSend qualified prospects to your Airtable base, with your approvalWhen people are qualified as Fit, an agent prepares a record for each one in your Airtable base and asks before writing, so your ops tracker stays current without copy-paste.Custom agentScheduledEvery morning, the people who turned high intent and who to contact firstEach weekday morning an agent collects the new Hot people from Intent, saves them to a list and tells you who to contact first and why.Custom agent
ScheduledRevive closed-lost deals with a reason to reconnectEvery Monday an agent reads deals lost more than 90 days ago in your CRM, checks what changed at each company, and drafts a reason to reconnect for the ones worth another look.Custom agent
ScheduledA weekly summary of replies and who to follow up withEvery Monday morning an agent reads last week’s replies, sums up what people said, lists who needs a follow-up and why, and emails it to you, with tasks in Todoist if you want them.Custom agent One agent per job, handing work to the next
Agents start each other through your lists and your inbox. Each one leaves a result you can check, and the launch waits for you.
- Research AgentSaves people who fit to a listOn your schedule
- Qualifier AgentTags Fit, Maybe or NoPeople added to a list
- Writer AgentWrites each first lineFit people reach its list
- Your campaignLinkedIn + email, stops on replyYou approve the launch
- Reply AgentDrafts the answer for youA labelled reply comes in
By hand, or with the Custom agent
| The job | By hand | Custom agent |
|---|---|---|
| Starting point | A blank automation canvas | A plain-language brief |
| Logic | Branches you wire by hand | Instructions and reusable skills |
| Your other tools | Glue code and API keys | Connected apps, or any MCP server or REST API |
| Control | All or nothing | Per-action permissions and approvals |
| Knowing what happened | Raw logs | A run page in plain words, credits per step |
Briefs that put the Custom agent to work today
Copy one into Penny, or use it as the agent’s instructions. Plain words are all it needs.
Every weekday at 9, list yesterday’s positive replies with the account owner and the next step, and email it to me.
When an import finishes, score the new people against my ideal customer and tell me the top ten.
When someone turns high intent, check Attio for an open deal, and if there is none, add them to “Hot this week”.
Every Friday, draw this week’s funnel as a diagram and summarise what changed.
Agents come with every Penny plan
Pick how much of the outbound day you want to hand over. AI work uses credits separately from the subscription, with a limit you set per run.
- 5 agents
- 5 connected apps
- Knowledge and skills
- Approvals and run limits
- 10 agents
- 10 connected apps
- Penny runs your outreach day
- Uses your own sending accounts
- 20 agents
- 25 connected apps
- Everything in Penny SDR
- One LinkedIn account and one mailbox included
What kind of agent can I build?
Any job you can describe to a teammate that uses your Prosyo data and connected apps: research, qualification, writing, list upkeep, reporting, CRM checks and follow-up drafts. Start from the Research, Qualifier, Writer, Reply Assistant or Intent Watcher template, or from scratch.
What can start an agent?
Run now, a schedule (every day, weekdays, or a day each week at a set time), or an event: people added to a list, a reply with a label, someone turning high intent, or an import finishing.
Can it use my own tools?
Yes. Tick any connected app from the directory. Owners and admins on Penny plans can also add a custom connector: an MCP server, or a REST API from its OpenAPI description. Reads run straight away; anything that changes data waits for approval.
How do I stop it doing too much?
Choose Review each step, Review changes or Run routine work, pick the exact actions it may repeat on its own, and set the most steps and credits one run may use. A run pauses at either limit and waits for you to continue.
How many agents can I have?
Five on Penny Agents, ten on Penny SDR and twenty on Penny SDR All-in. Up to two runs are active at a time in a workspace; others wait their turn.
Do I need to know how to code?
No. An agent is a brief in plain words, a trigger, the apps it may use and its approval rules. Even a custom connector only needs the service’s address or its OpenAPI description.
What are skills?
Short, reusable playbooks an agent loads when it needs them, such as Qualify against my ideal customer or Book the meeting. There are eight built in, you can copy one to make your own, and a workspace can keep up to fifty.
Can it use my files and website?
Yes. Add your deck, pricing, case studies and website to Knowledge. Agents use those facts and say which source an answer came from.
What happens if a run stops halfway?
Completed steps stay saved. A run that reached its step or credit limit waits with a Continue button, and you get a notification when a run finishes, stops or needs your approval.
What does a run cost?
Agent steps use AI credits like Penny questions. Paid actions such as enrichment, and charges from connected apps, are separate from the AI limit.
Your next 10 meetings are already on LinkedIn.
7 days free, no card. Pay later and change your mind? Full refund within 7 days.
- 7 days free
- No card required
- Cancel in two clicks
