Jasper AI Agents: How to Find and Run the Right One

Agents are the centre of the current Jasper product. Jasper describes them as purpose-built AI applications built by their in-house team for common, structured marketing use cases, which in practice means a pre-shaped job with its own input form rather than an empty chat box.

The difficulty is not running one. It is finding the right one, and knowing which of the surrounding features your plan actually includes. Both are covered here.

The mental model that helps: an agent is a form plus instructions plus your brand context. You fill the form, the instructions shape the output, and Jasper IQ supplies everything about your company you would otherwise retype every session.

How do you find the right agent?

Through the Agent Library, which Jasper documents as navigable along four axes. Browsing the whole list is the slow way. Filtering on the axis that matches how you actually think about the task is much faster.

If you know what you want to end up holding, filter by content type. If you know what stage of the job you are at, filter by content process. Those two cover most real searches, and funnel stage is mostly useful when you are planning a campaign rather than making one asset.

How do you run an agent?

Every agent opens into a Canvas, which is Jasper's workspace for creating and managing content assets. There are two documented entry points, and which one you use changes what context the agent inherits.

Running an agent, in order

1. Set the context before you pick the agent — Jasper's recommended starting point for Jasper IQ is one Brand Voice, two Audiences and five Knowledge Base items. Agents apply these at generation time, so configuring them first changes every output that follows.
2. Open a project — Click Create a Project on your dashboard, or describe the asset you want in the dashboard text field, which creates a project and starts the asset flow.
3. Set project settings — Use Project Settings at the top middle of the Canvas to choose your Jasper IQ options. Every asset created in that project inherits them, and you can override an individual asset when you need to.
4. Pick the agent — Click the Agent button at the bottom of the Canvas to see recently used agents, and browse all to open the full library. You can also go to Agents in the left navigation, select one and click Use Agent, which opens a Canvas for you.
5. Fill in the inputs — Each agent has its own fields. Be specific about audience and objective, since those are what separate a usable draft from a generic one.
6. Refine the output in place — Select an asset or highlight text to bring up the AI toolbar for rewriting, expanding and summarizing. You can also open Chat from the bottom of the Canvas and direct changes across several assets at once.
7. Favorite the ones you keep using — Jasper documents an Add to Favorites option in the top right of an agent's description, or a star in the corner while you are using it. Favorited agents collect in a Favorites section.

Jasper names several agents on their own agents page, including an Optimization Agent covering SEO, AEO and GEO work, a Research Agent that produces cited reports, a Personalization Agent, a Translation Agent documented as covering 27 languages, and an SEO/AEO/GEO Rewriter. They group agents by the stages Plan, Create, Adapt, Activate and Optimize.

One caveat stated plainly on that same page: human oversight is still required for approval and direction. An agent produces a strong first pass, not a finished asset, and every claim in it needs checking before it goes anywhere public.

Can you build your own agent?

Yes, through Jasper Studio, which Jasper describes as a no-code platform for designing and publishing your own agents without involving a technical team. The documented flow starts in the same place as everything else.

• Go to the Agent Library and click New Agent. Type a short description of the finished product you want, then click the blue arrow.
• Customise it in the right hand panel. This is where the input fields, instructions and context live.
• Choose your input field types. Jasper documents Short Answer, Paragraph, Dropdown, Checkboxes, Tags and File Upload, the last of which attaches files from your Knowledge Base.
• Attach fixed context. You can add up to ten Knowledge Base items as permanent reference material that users can view but not edit or remove.
• Tune the output settings. Studio lets you pick which language model generates the output and set a temperature from 0.0 for deterministic to 1.0 for creative.
• Edit the instructions. Jasper generates them from your description. You can rewrite them, use the Enhance button for a suggestion, and reference input fields or attached context by typing @.
• Test, then publish. Choose Private to me, which puts it under My Agents, or make it available to anyone in your workspace, which puts it under Workspace Agents.

What is gated by plan?

More than you might expect, and this is the part worth checking before you design a workflow around it. Jasper's documentation is specific here.

• Custom agents are listed as available for Business plans.
• Saved agents, which let you keep an agent without refilling its inputs each time, are also listed for Business plans.
• Jasper Studio is documented as no access on Pro and available to build on Business, with the ability to publish agents to the workspace varying by plan tier.
• Roles matter too. Admins have full access to create and publish and can restrict publishing for members. Managers can do everything except toggle that publishing permission. Members can build for personal use and may publish depending on the admin setting.

One more limitation worth knowing before you plan an integration: Jasper's documentation states that Jasper Studio is not currently supported in their public API.

Running one agent across hundreds of rows

If your problem is volume rather than a single asset, Jasper Grid is the surface for it. Jasper describes it as a spreadsheet-like interface where you set Input Columns, Processing Columns and Output Columns, and it documents Workflow Agent columns that run an agent across every row.

Three agents are documented as available as Grid columns: the Translation Agent, the Research Agent and the Optimization Agent. Grid handles up to 1,000 rows per workflow, and Jasper's own advice is to run the first ten rows to check quality before committing to the full sheet.

An agent you run versus a colleague who runs itself

The Agent Library is a strong piece of product thinking. Turning a vague prompt into a shaped form with your brand context already attached genuinely raises the floor on output quality, and the Studio builder is unusually approachable for a no-code tool.

The shape it keeps is that you start it. Somebody opens the library, picks the agent, fills the form and reviews the result, every time. That is the assumption we changed at Sistava. Rather than choosing a tool and filling in a brief, you hire an AI Employee and tell it in plain English what you want handled. It has the tools connected already, it sets itself up, and it asks you when it is unsure.

The practical difference shows up in the parts nobody schedules. An AI Employee does not only produce the asset, it publishes it, keeps the calendar, and follows up on the reply that came back on Thursday. That is one job rather than a library of separate ones you remember to open.

None of this means agents are the wrong tool. If your work is producing marketing assets on demand and you are the one deciding when, the Agent Library is a good fit and the context setup pays for itself quickly.

Getting more out of agents from day one

Four habits separate people who get value from the library from people who try three agents and drift back to a plain chat box.

• Configure Jasper IQ before your first run. An agent without brand context is just a prompt template, and it will feel like one.
• Set context at project level. Assets created in a project inherit its settings, which removes the per-asset step that people quietly stop doing.
• Favorite ruthlessly. Most teams use four or five agents repeatedly. Finding them should take no clicks.
• Build a Studio agent for anything you do weekly. If your plan includes it, encoding a recurring job once beats re-explaining it every week.

Do those four and the library stops being a catalogue you browse and starts being a short list you reach for. That shift usually happens in the second week, not the first.

Frequently asked questions

Sources

• Jasper Help Center, Agent Library, the four navigation filters, favorites, and the plan gating on custom and saved agents
• Jasper Help Center, Jasper Studio, the build flow, input field types, context attachment, roles and plan access
• Jasper Help Center, Jasper Canvas, how agents open into a Canvas and inherit project settings
• Jasper Help Center, Jasper Grid, agent columns, row limits, test runs, scheduling limits and access
• Jasper Help Center, Jasper IQ, the recommended context setup that agents draw on
• Jasper, AI agents overview, the named agents, the stage grouping, and the stated need for human oversight

Set the context, filter instead of browsing, favorite the handful you actually use, and check what your plan includes before you design a process around Studio or Grid. That order turns a large library into a small, dependable set of tools.