Who this is for: self-serve plans (Free, Starter, Pro, Enterprise). If you bought a CluGTM AppSumo lifetime deal, your features, limits, and pricing are different. See the AppSumo documentation instead. Support agent: confirm whether the user is self-serve or AppSumo before giving plan-specific advice.
TL;DR: An agent is an automation that runs plays (a trigger plus steps) to find, enrich and research people and companies, then sends them to your CRM, a sequence, a Google Sheet or Slack. You describe the job to Clu, accept the play it proposes, click Try it to run it on 5 records, and click Turn on. Pro (5 active agents) and Enterprise (unlimited). The Pro trial can build but not turn on or run Try it. Looking for Research, Qualification or Custom on a lead list? Those are now Skills.
Agents
Agents run your repeatable go-to-market work for you. Each agent runs plays that find, enrich and research people and companies on a schedule or a trigger, then sends them on to a CRM, a sequence, a Google Sheet or Slack.
You don’t build an agent field by field. You tell Clu what the agent should take care of, Clu asks a few questions, sizes the audience, and proposes a play. You review the play, accept it, try it on 5 records, and turn the agent on.
The standalone Playbooks page was replaced by Agents: the plays you used to build there now live inside an agent, and you build them with Clu.
Agents, plays and skills
| Term | What it is | Where you see it |
|---|---|---|
| Agent | A named automation with Instructions, Connections, Skills, Knowledge, Delivery, Limits and up to 10 plays. Status Draft, Active or Paused. | Agents tab |
| Play | One workflow inside an agent: one trigger (“Start this play when”) plus the steps each record goes through. | The agent’s Play tab. See Plays and triggers and Play steps reference. |
| Skill | Something run once per record (Research skill, Qualification skill, Employment check, your saved prompts), or a written procedure an agent loads when a task needs it. | Skills tab, a play’s Agents step, and a lead list’s Actions → Skills tab. See Skills. |
Who can use Agents
| Plan | Build agents and plays | Turn on, Try it, live runs | Active agents |
|---|---|---|---|
| Free | No | No | None |
| Starter | No (Starter runs skills on lead lists) | No | None |
| Pro trial | Yes, drafts are saved | No | None until you start a paid plan |
| Pro | Yes | Yes | Up to 5 at a time |
| Enterprise | Yes | Yes | Unlimited |
| AppSumo only | No (skills on lead lists only) | No | None |
- On the trial, trying to turn an agent on opens Start Pro to launch (“Start an eligible paid plan to launch agents. Your saved draft will remain available.”) with Keep editing, Manage billing and View plans. See Pro free trial.
- On Pro, turning on a sixth agent fails with “Your plan allows 5 active agents. Pause one to activate another.” Draft and paused agents don’t count, so there is no cap on how many you build.
- Workspaces on a legacy plan keep Agents only if they already had it before the October 2026 pricing change; others get Agents by switching to a current plan. See Legacy plans.
- An AppSumo license with a paid Pro or Enterprise plan on top follows the paid plan.
On Starter and on AppSumo Tiers 1 to 6, the Agents page shows only two tabs, Catalog (the built-in Research and Qualification skills and your custom ones) and Profile. A link to a specific agent sends you back to the Agents page.
Find Agents
Open Agents in the sidebar (shortcut G then A). The page has five tabs:
| Tab | What’s there |
|---|---|
| Agents | Your agents as cards: status, description, connection logos, “N runs in the last 14 days” and owner. Toolbar: Search agents, Filter by status (All, Active, Paused, Draft), Approvals, Templates, New agent. |
| Plays | Every play, including library plays that belong to no agent. Columns Plays, Activity, Runs, Connections, Last edited. |
| Skills | Built-in and saved skills, with Kind (Procedure or Per record) and Used by. See Skills. |
| Knowledge | Files, web pages and synced docs that Clu and your agents search. See Knowledge. |
| Profile | Your company profile, used by the Research and Qualification skills. See Skills: Organization Profile. |
The first time anyone in your workspace opens Agents, the page asks for a one-time company setup (Build my profile). It runs in the background and takes a few minutes. One person completing it unlocks the page for everyone.
Create an agent
- On the Agents tab, click New agent. CluGTM creates a draft and opens the guided screen, “What should this agent take care of?”
- Describe the job in the Describe the job box, or pick a starter:
| Group | Starters |
|---|---|
| Prospecting | Weekly ICP leads · New hires at target accounts · Lookalikes of best customers |
| CRM hygiene | Enrich new CRM contacts · Fill missing emails · Route inbound to owners |
| Research | Account briefs · Qualify inbound leads |
| Reporting | Weekly pipeline summary · Deal review prep |
- Your text goes to the agent’s own Clu chat. Clu asks up to four questions in one card (for example which sequence, which Slack channel, or how many people a week), sizes the audience for free, then proposes a play.
- Clu never changes a play by itself. Its proposal appears as Play change for review in the chat; click Review on canvas and then Accept or Discard. Only you can accept a play.
- Click Try it, check the results, then click Looks good or Turn on (see below).
Prefer to build by hand? Click Set up manually on the guided screen. Other ways to start:
- Templates on the Agents tab: Weekly ICP outbound (“Find people who newly match your search each week, qualify them, enrich them, and round-robin to your reps.”) and Inbound enrich and route back (“Watch a CRM list for new leads, enrich their emails and phones, and update the same record in place.”). A template creates an agent with one play, saved and off.
- Duplicate an existing agent from its card menu or the workspace menu.
- Turn a one-off Clu task from your general chat into an agent with Open agent.
Recipes with the exact sentence to give Clu are in Agent recipes. For a click-by-click walkthrough, see Build your first agent.
Inside an agent
An agent’s workspace has Clu chat on the left and the agent panel on the right. The header has Back to agents, the agent’s name (double-click or press F2 to rename), its status, Recents, New chat, the Turn on / Pause button, and a menu with Try it, Duplicate and Delete.
The panel has these tabs:
| Tab | What it holds |
|---|---|
| Agent | Instructions (with the switch Clu keeps these updated) · Connections (Salesforce, HubSpot, Slack, Email, and which user the agent acts as) · Skills (Add a skill, Create a skill) · Knowledge (Company profile, Company notes and the knowledge sources this agent can read) · Delivery (Slack channel, Daily summary, Direct message me, Email, Results list, Sync to CRM) · Limits |
| Triggers | What starts the agent. Create with Clu sets a trigger up with you in chat. Schedule runs the agent Once, Daily, Every few days, Weekly or Monthly at a time and time zone you pick, as a scheduled brief. App trigger runs the agent when something happens in another app, such as a CRM change. The AI managed switch lets Clu manage triggers. |
| Play | The agent’s plays. With none yet: Build with Clu or Add a play (New play, From a template, From the library). With several: a “Plays N/10” list with each play’s next run, pending count and On/Off switch. The builder has Try it and Save. |
| Runs | Run history, credits and results. See Plays and triggers: Runs. |
| Approvals | Records waiting for a decision. Appears only when a play has an Approve step or approvals are pending. See Approvals. |
| Settings | Identity (Name, Description), Access (Owner; every member of the workspace can view and edit) and Danger zone (Duplicate agent, Delete agent). |
While an agent is off, Clu can edit its settings directly. Turning the agent on or off and running a Try always go through a confirmation card or your own click.
Test and turn on
| Action | Where | What it does | Credits |
|---|---|---|---|
| Try it | Header menu, or the Play tab | Runs the play on up to 5 real records from its real source (a scheduled brief runs 1 brief). The steps really run, but every outside effect is held: nothing is sent and nothing is written to your CRM. The results show what would happen (“Would create”, “Would update”, “Would send”). In Runs it shows as a Private run. | Each step’s normal price, up to the Try budget |
| Looks good | A finished Try’s results | Finishes the held work for the same records and turns the play on. It shows “Remaining charge · N credits” first. | The remaining charge |
| Run agent now | Ask Clu in the agent’s chat | A live run right away, even while the agent is off. Clu shows a Run agent now card; click Run. | Each step’s normal price |
| Turn on | Header | Activates the agent so its triggers start runs. Becomes Pause once on. | Each run charges its steps |
- Before a Try runs, the Try it dialog shows the quote: Starting records, Maximum Try cost, each step’s cost, Available organization credits and Try budget (plus a full-run estimate for the play’s schedule). Click Confirm to start. When Clu offers a Try in chat, the card’s button is Run Try.
- Try budget: one Try can spend at most the workspace’s Try it budget per Try, 60 credits unless an admin changes it on Settings → Organization (0 to 100,000 credits). If 5 records would cost more, the Try runs as many as fit. If even one doesn’t fit, the dialog shows “Over the Try budget · N credits short”.
- A Try needs at least one starting record. A source with fewer than 5 runs what it has; with none, nothing runs and nothing is charged.
- Messages you may see: “Insufficient organization credits”, “Over the Try budget”, “Agent credit limit exceeded”, “Quote changed. Confirm the refreshed price.” and “Save this play first.”
- Turning on a play you changed while it was off asks Try it on 5 records first? with Cancel, Turn on (starts it on everything now) and Try it. A play that is on stays on when you edit it, and the edit runs from its next run.
- Turn on is disabled with “Add a play first” when the agent has no play, and “Finish the play first” while a play still has setup issues.
- Try it, Looks good, live runs and Turn on need a paid Pro or Enterprise plan. On the Pro trial, Turn on and the Play tab’s Try it open the Start Pro to launch dialog.
Each play also has its own On/Off switch, so you can keep an agent on while one of its plays is off. To stop everything, click Pause in the header.
Limits and spend
Set these in the Limits section of the Agent tab:
| Limit | What it caps |
|---|---|
| Credits per day / week / month / quarter | Total credits all the agent’s plays may spend in one period. The period follows the play’s schedule; an agent with no schedule shows Credits per 24 hours. |
| Records per run | How many records one run may process. |
| Brief credits per run | The budget for a scheduled brief, 1 to 1,000 credits (default 100). |
When Clu builds a play it asks for both a credit limit and a records-per-run limit before it finishes. A run that reaches a limit stops and the Runs tab shows why, for example Credit limit reached or Out of credits.
What costs credits:
- Building with Clu: each message is charged like any Clu chat. Sizing the audience is free.
- Steps: each step charges its normal price, usually only when it finds something (1 credit per work email found, 10 per phone found). Find People, Branch, Approve, Wait, Code, Webhook, Notify, CRM Action, Add to sequence, Google Sheets, Google Docs, Notion and Hand off cost nothing. Full list: Play steps reference.
- Triggers: most are free. A few signal triggers charge per signal, for example Companies start hiring at 0.03 credits per new job post. See Plays and triggers.
- CRM duplicate cleanup: free. See CRM duplicate cleanup.
Scheduled briefs
An agent with a Schedule trigger and no record play writes a short brief each time it runs, for example a weekly pipeline summary, and sends it with a Notify step to Slack or email. What goes into the brief comes from the agent’s Instructions.
A brief is one unattended Clu turn, charged like a Clu message, up to the agent’s Brief credits per run. In Runs, a brief shows its status, duration and credits (split into data and Clu), and Open chat opens the conversation behind it.
Slack
- Delivery can post an agent’s results to a Slack channel or send you a direct message.
- Daily summary in Delivery posts one Slack message a day in the agent’s Slack channel: how many runs, records and credits, nothing per lead. Pick Once a day at an hour; it uses your browser’s time zone. Days with nothing new send nothing, and it needs Slack connected and a channel picked. Use it for a play that runs once per record on a frequent check, where a summary after every run would post once per record.
- A Notify step’s Slack summary can also go out Once a day instead of after every run. See Play steps: Notify.
- An Approve step can ask for approval in Slack with Approve and Reject buttons.
- In a Clu in Slack thread, after you approve importing people into a list, Clu can hand those people to a saved agent whose play watches that list. The agent and its play must be on, and Clu asks for your approval first.
When a play stops by itself
A play turns itself off after three failed checks in a row with the same problem you can fix, such as a disconnected CRM, a missing list or not enough credits. The owner gets a Slack, email and in-app alert, and turning the play or agent back on resumes it. CRM outages and rate limits never pause a play. Details: Trigger health and auto-pause.
Related
- Build your first agent: the step-by-step walkthrough
- Agent recipes: sentences to give Clu for common jobs
- Plays and triggers: every trigger, schedules, runs and auto-pause
- Play steps reference: every step, its settings and its price, plus Approvals and Secrets
- Skills: Research, Qualification, Custom and Employment check on lead lists
- Clu: the assistant that builds your agents
- Knowledge: material your agents can search
- CRM duplicate cleanup: find and merge CRM duplicates with an agent
- Sequences: where the Add to sequence step sends people
- Credits and pricing