On this page
What TextCortex does
TextCortex combines custom agents, knowledge bases and search tools. For students, that can connect repeated reading questions with a defined source collection. The useful question is whether you can recover and verify the evidence behind an answer. This review is based on official documentation, without a completed accuracy benchmark or hands-on claim about academic results.
Where it fits, and where it falls short
Useful strengths
- Knowledge bases can organize multiple files for focused questions and comparisons.
- Agent instructions and default sources can support a consistent study routine.
- Search tools distinguish web discovery from work with selected documents.
Limits to consider
- Retrieval selects relevant sections; it does not mean the entire collection was read for every answer.
- Higher-compute models and extended agent runs draw on a usage budget.
- Source references require checking, and connected information may be a synced copy.
A suggested TextCortex workflow
View the workflow diagram
- 1
Choose a small, permitted reading set and keep full source details.
- 2
Create a focused knowledge base and define an agent that identifies missing evidence.
- 3
Ask one comparison question, inspect cited passages and correct your notes.
- 4
Write your own argument and check usage before expanding the workflow.
Build a focused reading collection
The knowledge-base guide describes uploading documents, selecting context in chat and retrieving sections relevant to a question. It supports sources such as PDFs, Word documents, spreadsheets and public URLs. Start with one assignment rather than mixing unrelated courses into a single repository; the guide itself recommends grouping collections by purpose.
Retrieval is selective. A broad prompt about everything in a folder can miss qualifications that matter to your argument. Begin with a specific definition or method question, then compare documents in smaller steps. Preserve the original file and publication details outside generated notes so that a later answer does not become your only record of the evidence.
Supporting documentation: [1]
Give a study agent a narrow role
TextCortex documents configurable agent backgrounds, behavioral rules, default knowledge bases and tools. For a reading assistant, specify the task and the allowed material: distinguish the authors' findings from your interpretation, identify the source for a claim and report when the selected material is insufficient. These are suggested instructions, not guarantees that the model will follow every rule.
Imagine a teacher-training assignment comparing formative feedback in two classroom studies. Upload the permitted papers and your assignment rubric. Ask for a table separating participants, feedback timing, outcome measure and stated limitation. Then request two questions that a student should answer before treating the findings as applicable to a different classroom. This hypothetical exercise has not been tested here.
Supporting documentation: [2]
Verify the passage before writing the conclusion
For that example, open the passage behind each table entry. Improved immediate performance and improved retention after a delay are different outcomes. Correct any collapsed distinction in your notes before developing a thesis. Record which details the assistant failed to recover, alongside useful findings; polished wording alone gives little information about evidential quality.
The web-search guide describes clickable references and explicitly advises fact-checking. A link to a real paper is only the beginning: inspect whether the relevant result concerns the right population, intervention and measurement. Keep missing or inaccessible evidence marked as unresolved rather than asking the system to fill a gap with a plausible explanation.
Scholar search is a discovery tool, not a completeness guarantee
The search documentation includes Google Scholar among selectable sources, alongside general web and other modes. TextCortex therefore offers more than a private-file chatbot. Use this to discover candidate readings, then check publication details and obtain the full text through an appropriate source. Record the search terms and selection decisions yourself.
The same guide says web sources must be publicly accessible, an imported URL is not updated in real time, and a homepage request does not automatically collect its entire website. A generated research report does not establish that all relevant scholarship was searched. A systematic review still needs its defined databases, reproducible queries and screening process.
Supporting documentation: [3]
Budget for the work an agent performs
The usage guide describes a default compute wallet covering model questions and longer agentic work. When it is exhausted, the documented fallback is standard models and default settings. Optional Extra Usage requires activation and is billed using underlying model cost plus a platform markup, with administrator spending controls. This differs from treating every chat message as an identical creation.
Begin with a representative reading task and inspect the usage display before repeating it across many files. Check whether Extra Usage is enabled and which caps apply. A subscription price alone does not describe a heavy research workflow's resource allowance. Avoid carrying old daily-creation figures from extension listings into a purchasing decision without confirming the current account.
Separate copied knowledge from live connected actions
The connector guide distinguishes knowledge integrations that sync copies from MCP connectors that access services live. Connectors can also write when their tools permit it. Review the selected tools and permissions before connecting a shared course drive. The freshness of a synced knowledge collection depends on its synchronization, rather than simply the original document's modification date.
The current pricing FAQ states that data is not used for model training, while a still-accessible 2023 data article describes different general-usage terms and an opt-out. Resolve the applicable terms for your account before uploading sensitive work; this review does not treat those pages as a uniform retention promise. Compare NotebookLM for a source collection or Consensus for scholarly discovery when those narrower jobs dominate.
Pricing, free access and plan checks
The pricing page displayed Premium at $29.99 per user monthly or $23.99 per user per month billed annually, excluding VAT. Enterprise is custom-priced; a 14-day trial is advertised. Model usage and optional Extra Usage require separate attention. Confirm the billing total, trial conditions and account allowance before subscribing.
Confirm billing period, renewal terms, limits and regional availability on the provider’s current plan page before paying.
Questions about TextCortex
Is TextCortex free?
The current pricing page advertises a 14-day trial. Confirm its conditions and the allowance shown in your account; older free daily-creation figures are not used as a current entitlement here.
Can TextCortex search scholarly sources?
Its search guide documents a Google Scholar source option. That helps discovery but does not demonstrate exhaustive coverage or replace checking the original publication.
Do knowledge bases read every document for each answer?
The guide describes retrieving relevant sections. Ask focused questions and verify the supporting passages, especially when comparing methods or conflicting findings.
Sources and research notes
This guide synthesizes official product information with our editorial assessment of academic use. Suggested exercises are examples you can repeat; they are not results of tests we performed. We do not infer accuracy from a provider’s marketing claims.
- Knowledge-base retrieval, file types and limits
- Custom agent behavior, rules and tools
- Web and Google Scholar search, citations and limitations
- Current pricing, trial and data-training statements
- Model usage wallets, fallbacks and Extra Usage
- MCP connectors versus synchronized knowledge bases
- Older data-handling article dated July 25, 2023
Found a changed feature or a factual error? Send the exact page and an official source to info@ai-essay.com. Read our editorial policy for how we handle corrections.