Manus Now Suggests Its Own Research Sources so You Don't Have To
Manus adds smart source suggestions to projects, letting the agent learn which web pages to reference across future tasks — no URL hunting required.
- New feature: Manus can now suggest relevant web pages for a project, which are then automatically referenced in all future tasks.
- How it works: Go to Project → Files & Sources, describe what you need, pick from Manus's suggestions, and those sources persist across tasks.
- Problem solved: Eliminates manual URL hunting and ensures the agent consistently uses the right sources, including paywalled ones you have access to.
- Available now: The feature is live for all users on all plans, including the free tier (300 daily credits).
- Pricing: Free plan exists; paid plans run $20–$200/month based on credit volume, with a 17% annual discount.
- Limitations remain: Mid-task steering is still limited, long tasks can crash and burn credits, and deep coding work is better handled by dedicated agents.
Manus, the autonomous AI agent now part of Meta, just shipped a small but meaningful quality-of-life update to its Projects feature. Instead of manually tracking down and pasting exact URLs for the agent to reference, you can now describe what kind of sources you need and let Manus suggest the relevant pages itself. Once you pick the ones you want, those sources are pinned and automatically referenced in every future task inside that project.
The problem it actually solves
Manus is built around long-running, multi-step research tasks. It uses a multi-agent architecture where a central controller breaks your goal into sub-tasks and delegates them to specialized sub-agents for web browsing, code execution, data analysis, and file management. The catch has always been source quality. A significant chunk of the research-grade web , academic databases, premium news, gated SaaS pricing pages , is invisible to Manus, and you will not realize data is missing until you check manually.
The new source suggestion feature is a direct answer to that gap. Rather than hoping the agent finds the right pages on its own, you now have a structured way to tell it exactly where to look , and that guidance persists across the life of a project.
How to use it
The workflow is straightforward:
- Open a Project and navigate to Files & Sources
- Describe what kind of information or sources you need
- Manus suggests relevant pages
- Pick the ones you want to keep
- Those pages are now referenced automatically in all future tasks in that project
This is available to all Manus users right now, on any existing plan.
Why this matters for research workflows
Manus is strongest when a task has many steps, requires synthesis, and benefits from a deliverable at the end. The typical use case is something like a competitive analysis, a market research report, or a technical literature review , tasks where you want to point the agent at a specific set of authoritative sources and have it work from those consistently.
Before this update, the friction was real. Manus allows stepping in mid-task and tweaking or redirecting actions, and it gives users confidence to intervene and select more relevant sources when needed. But doing that manually, task by task, was tedious. Now the source selection happens once, upfront, at the project level.
This is especially useful when you have access to paywalled or authenticated sources. Manus can assist within pages you have permission to access, including premium resources like Crunchbase, PitchBook, SimilarWeb, Financial Times, Semrush, and Ahrefs , any tool or source you are subscribed to. Pinning those pages to a project means the agent knows to use your premium data, not just whatever it can find publicly.
Where Manus still falls short
The source suggestion feature helps, but it does not fix Manus's deeper reliability issues. Users frequently encounter server overloaded messages during peak hours, and long-running tasks (20+ minutes) sometimes crash mid-execution with no recovery , and you lose the credits spent on the failed task. There is also the steering problem: once Manus starts executing, you have minimal ability to redirect it; if the agent misinterprets your goal, your options are to let it finish and waste credits, or cancel and start over.
For deep coding work, dedicated tools are still the better call. For multi-file refactors, IDE-grade context, and PR-style workflows, dedicated coding agents like Claude Code, Cursor, and Cline remain stronger.
Pricing at a glance
Manus offers a free plan with 300 daily credits, a $20/month Pro tier with 4,000 credits, a $40/month tier with 8,000 credits, and a $200/month tier with 40,000 credits. Annual billing saves 17%. The source suggestion feature is available across all plans. The catch is the credit system: a single deep-research task can quietly burn through 900 to 1,000+ credits of your monthly allocation, and Manus does not tell you the cost before you hit go.
The source suggestion update is a minor feature on paper, but it reflects a more important design shift: moving Manus from a one-shot task runner toward something closer to a persistent research collaborator. For anyone already using Manus for recurring research projects, it is worth trying immediately. For those evaluating it for the first time, the free tier gives you enough credits to see it in action without committing.
Comments
