Skip to content
WorkspacesHow it worksSourcesDeep AnalysisPricingFAQ
Source-grounded workspace for serious document work

Turn your documents into a body of knowledge.

Bring the files that define a project into one persistent workspace. Ask across the collection or focus on one source, inspect the evidence behind the answer, and return without rebuilding the context.

Persistentfiles · context · conversations
Traceableclaims · sources · excerpts
AI Papers
ACROSS AI PAPERS · 6 SOURCES

What tradeoffs do these documents identify between model quality, latency and inference cost?

The sources broadly agree that higher model quality can increase latency and cost. Several recommend routing routine requests to smaller models while reserving larger models for tasks requiring deeper reasoning. [1] [2]

[1]compute-market-analysis.pdf§3.2
[2]model-routing-strategies.pdf§1.4
The real problem

The documents exist. The understanding is still fragmented.

01

The evidence lives in too many places

Files sit across folders, inboxes, drives and tabs. The information is available, but the project has no single working context.

02

The question crosses document boundaries

The answer depends on several sources, but most workflows force you to inspect and summarize them one at a time.

03

The answer loses its provenance

A useful paragraph gets copied into a note or presentation, while the exact source and supporting excerpt disappear.

04

The work resets between sessions

Previous questions, instructions and conclusions become another trail you must reconstruct before the next round of work can begin.

The mechanism

One workspace holds the entire line of inquiry.

Search+ keeps the project’s files, standing instructions, conversations and source evidence inside one persistent environment. Every new question begins with the context the workspace has already accumulated.

A

Files

Reports, papers, contracts, notes, spreadsheets and supported web pages.

B

Workspace Context

Set priorities, definitions and constraints once, applied to every question.

C

Conversations

Keep separate lines of inquiry and reopen them when the work continues.

D

Scope

Ask across the full collection, or deliberately narrow to one source.

AI Papers
6 DOCUMENTS
PDFinference-systems-overview
PDFmodel-routing-strategies
PDFretrieval-evaluation
DOCagent-governance-notes
PDFcompute-market-analysis
XLSresearch-summary
Workspace Context

“Prioritize technical findings, distinguish evidence from interpretation, preserve source citations and make important limitations explicit.”

The life of a workspace

Watch a folder become a working body of knowledge.

The documents arrive first. Then the workspace gains instructions, questions, evidence, conversations and reusable analysis.

Search+ · AI PapersStep 1 / 8
STEP 1 · CREATE

Create a permanent home for the project

Name the workspace once. It becomes the continuing environment for the documents, questions and conclusions surrounding that body of work.

+
AI Papers
NEW WORKSPACE · 0 DOCUMENTS · SAVED
STEP 2 · ADD DOCUMENTS

Add the source material

Bring in PDFs, documents, spreadsheets, text files or supported web pages. Search+ processes each source so it can participate in the workspace.

PDFinference-systems-overview.pdfINDEXED
PDFmodel-routing-strategies.pdfINDEXED
PDFretrieval-evaluation.pdfINDEXED
DOCagent-governance-notes.docxINDEXED
PDFcompute-market-analysis.pdfINDEXED
XLSresearch-summary.xlsxINDEXED
STEP 3 · CONTEXT

Set the rules for the work

Add standing instructions that should shape every answer: priorities, terminology, exclusions, evidence standards or required output conventions.

Workspace Context · applies to every question

“Prioritize technical findings, distinguish evidence from speculation, preserve source citations, and explain important limitations.”

CONTEXT ACTIVE
STEP 4 · ASK ACROSS

Ask a question no single file can answer

Search across the eligible collection to compare claims, identify patterns and synthesize evidence distributed across several sources.

ACROSS AI PAPERS · 6 DOCUMENTS

What are the major tradeoffs these documents identify between model quality, latency, and inference cost?

Higher quality trades against latency and per-query cost. Routing routine requests to smaller models recovers most of the savings [1] with no measurable quality drop on routine tasks [2].

STEP 5 · FOCUS

Narrow the scope when precision matters

Pin the conversation to one document when the question requires a focused, source-specific answer.

PDFmodel-routing-strategies.pdf · SINGLE-DOCUMENT

What routing strategy does this document recommend, and under what conditions?

It recommends confidence-based routing: send routine extraction and classification to a small model, and escalate to a larger model only when confidence is low or the task is reasoning-heavy [1].

STEP 6 · SOURCES

Inspect the evidence behind the claim

Open a citation to see the source, available location metadata and the exact excerpt supporting the answer.

…recovers most of the savings [1] with no measurable quality drop [2].

[1]compute-market-analysis.pdf§3.2
“Marginal quality gains above a threshold carry disproportionate cost. Selective routing preserved 96% of task success while cutting spend materially.”
[2]model-routing-strategies.pdf§1.4
“For routine classification and extraction, smaller models matched larger ones within noise. Reserve premium capacity for reasoning-heavy prompts.”
STEP 7 · RETURN

Return without reconstructing the work

Reopen an earlier thread with its workspace, scope and conversational context intact.

Tradeoffs: quality vs latency vs cost
Across AI Papers · 3 days ago
REOPEN →
Routing recommendation
model-routing-strategies.pdf · 5 days ago
REOPEN →
Retrieval evaluation notes
Across AI Papers · last week
REOPEN →
STEP 8 · DEEP ANALYSIS

Turn the collection into a structured analysis

Compare, synthesize or extract across selected sources and preserve the result as part of the continuing workspace.

Inference demand: comparison6 SOURCES
AGREEMENT

All six expect demand to rise; the split is on how steeply.

DISAGREEMENT

Two treat capacity as binding; others emphasize latency.

DOCUMENT
OUTLOOK
DRIVER
compute-market
Steep
Capacity
inference-systems
Moderate
Latency
model-routing
Moderate
Routing mix

A workspace is not an upload folder. It is a persistent body of files, context, conversations, evidence, and reusable analysis.

Control the scope

Ask the collection. Interrogate the source.

ACROSS AI PAPERS · 6 DOCUMENTS

“What tradeoffs do these sources identify between model quality, latency and inference cost?”

Search across every eligible source when the answer depends on comparison, synthesis or evidence distributed throughout the collection.

model-routing-strategies.pdf · SINGLE-DOC

“What routing strategy does this document recommend, and under which conditions?”

Narrow the conversation to a specific document when you need an answer grounded only in that file.

Source grounding

Keep the answer attached to the evidence.

Inline citations connect important claims to the source material behind them. Open a citation to inspect the document, available location metadata and the supporting excerpt.

Search+ shows sections and page references when the underlying source provides them. It does not manufacture location data that the document does not contain.

Selective routing preserved task success while cutting spend; smaller models matched larger ones on routine tasks [1]; smaller models matched larger ones on routine tasks [2].

[1]compute-market-analysis.pdf§3.2
“Marginal quality gains above a threshold carry disproportionate cost. Selective routing preserved 96% of task success while cutting spend materially.”
[2]model-routing-strategies.pdf§1.4
“For routine classification and extraction, smaller models matched larger ones within noise. Reserve premium capacity for reasoning-heavy prompts.”
Deep Analysis

When the question spans documents, the result should show its work.

Select the relevant sources and choose an analysis mode. Search+ compares, synthesizes or extracts across the collection and returns a structured result connected to the underlying evidence.

Task

Compare the six sources’ assumptions about inference demand, identify agreements and contradictions, and produce a sourced synthesis.

COMPARESYNTHESIZEEXTRACT
Reviewing the selected sources
Comparing claims and findings
Organizing the evidence
Preparing the structured result
Inference demand: comparison6 SOURCES
DOCUMENT
OUTLOOK
DRIVER
compute-market
Steep
Capacity
inference-systems
Moderate
Latency
model-routing
Moderate
Routing mix
AGREEMENT

All six expect rising demand; the split is on how steeply.

DISAGREEMENT

Two treat capacity as binding; others emphasize latency.

Every important conclusion remains connected to its source. Continue the inquiry from the result.

Applications

Built for questions that do not fit inside one document.

Review research literature

Compare methods, findings and limitations across a reading list.

Compare reports

Surface agreements, differences and quiet contradictions between sources.

Work through board or diligence materials

Question decks, memos, models and supporting documents inside one workspace.

Trace a claim

Follow a statement back to the excerpt that supports it.

Analyze interviews

Find recurring themes and disagreements across transcripts.

Extract structured findings

Pull the same fields, claims or evidence types from several documents.

Build an evidence brief

Organize the strongest findings and the sources behind them before making a recommendation.

Investigate one source

Narrow the scope for precise questions about a specific document.

Source ingestion

Bring the documents the work actually depends on.

PDFDOCXXLSXTXTURL
PDFretrieval-evaluation.pdfINDEXED
DOCagent-governance-notes.docxPROCESSING
TXTmeeting-notes.txtQUEUED
Recurring value

Each return should start further ahead.

Files and standing context remain attached to the project. Nothing to reconstruct next week.

Reopen an earlier conversation and continue from the reasoning already established.

New sources extend the workspace while previous questions and conclusions remain available.

Deep Analysis results become reusable work products rather than disposable responses.

Pricing

Choose the capacity your work requires.

Use Search+ as a personal knowledge workspace, collaborate with a small team or secure early lifetime access without a recurring subscription.

Pro
$9.99/month
Billed monthly.

  • +600 pages included per month
  • +$0.02 per additional page
  • +Unlimited Workspaces
  • +Chat with documents
  • +Deep Analysis (beta)
  • +OCR
  • +No document size limit
  • +Custom Deep Analysis Templates
  • +Early access to new features
Start with Pro
Team
$39.99/month
For a team of 3, billed monthly.

  • +Team of 3 users included: the account owner plus 2 additional users
  • +Role-based access control
  • +3,000 pages included per month
  • +Each additional user gets 250 pages
  • +$0.01 per additional page
  • +Unlimited Workspaces
  • +Chat with documents
  • +Deep Analysis (beta)
  • +OCR
  • +No document size limit
  • +Custom Deep Analysis Templates
  • +Early access to new features
Start a Team workspace
Early Access
Lifetime Access
$149one-time
One-time payment.

  • +Up to 100 chats per month
  • +Up to 10GB storage
  • +Access to all core features
  • +No recurring subscription
  • +Pay-as-you-go for additional usage
Get Lifetime Access

Monthly usage limits and pay-as-you-go charges still apply.

Trust

Serious document work needs visible evidence, and visible limits.

Inspect the source

Citations open to the source material behind the answer, not merely a document title.

Control the scope

Search+ makes it clear whether a response draws from the workspace or one selected source.

Keep judgment human

Search+ helps you investigate and organize evidence. It does not guarantee that an answer is complete, correct or appropriate for a consequential decision.

FAQ

Questions, answered.

What is Search+?+
A persistent, source-grounded workspace for working with a collection of documents. It keeps the files, standing context, conversations, question scope and supporting evidence connected throughout a project.
How is Search+ different from a basic document chatbot?+
A basic document chatbot is usually optimized to answer an isolated question. Search+ is organized around a continuing workspace: multiple sources, persistent instructions, separate conversation threads, explicit query scope, inspectable citations and structured Deep Analysis.
Can I ask questions across several documents?+
Yes. Workspace-level questions can use every eligible source in the collection, making it possible to compare findings, identify patterns and answer questions that span several documents.
Can I limit a question to one document?+
Yes. Select a single source when the answer should be grounded only in that file. Search+ keeps the active scope visible throughout the conversation.
Does Search+ show where an answer came from?+
Search+ provides inline source references for grounded answers. Open a citation to inspect the supporting excerpt and any section or page metadata available from the original document.
What does Workspace Context do?+
Workspace Context stores standing instructions for the project. You can define priorities, terminology, constraints and evidence standards once, and Search+ applies them across future questions in that workspace.
What is Deep Analysis?+
Designed for tasks that require structured work across several sources. Choose a source set and an analysis mode, such as Compare, Synthesize or Extract, and Search+ prepares a sourced result you can continue working from.
Will my workspaces and conversations be saved?+
Yes. Workspaces and conversations persist so you can return to an earlier line of inquiry instead of reconstructing the project every time.
Which file types does Search+ support?+
PDF, DOCX, XLSX and TXT files, as well as supported pages added from a URL. OCR is available for documents that require text recognition.
Is there a document-size limit?+
Search+ does not impose a plan-level document-size limit. Processing time can vary depending on the file’s size, structure, image content and OCR requirements.
Can Search+ guarantee that an answer is correct?+
No. AI-generated answers can be incomplete or inaccurate. Search+ makes the underlying evidence easier to inspect, but you remain responsible for reviewing important claims before relying on them.
Begin

Build the body of knowledge your next decision depends on.

Start with the documents already defining the work. Keep the questions, context and evidence connected as the project evolves.