Use case · Research

Compose a specialist agent for every layer of research.

One agent hunts primary sources, another triangulates across them, another verifies the numbers, another synthesizes the memo or deck — every agent loaded with your methodology, every claim traceable to its source. Works for any research domain: legal, financial, real estate, market intelligence, scientific.

Best fit for · Research teams shipping memos, decks, or briefs across legal, financial, market, real-estate, or technical domains weekly
Request beta accessSee an example pack
vs Perplexity Computer →
packwolf.app · Research workspace
Live
Project · Q2 competitive analysis6 SUB-QUESTIONS · 28 CLAIMS · METHODOLOGY-LOADEDSOURCESCLAIMSMEMORLMANAGERResearch LeadSHSource HunterSOURCES: 42TRTriangulatorVERIFYING 18QAQuantRE-RUNNING 6SYSynthesizerDRAFTING MEMOTO REVIEWMemo · pendingSCOPE · DECOMPOSEDHow are competitors moving in Q2?OWNED BY: RESEARCH LEAD1.1Competitor pricing structuresSH7 CLAIMSVERIFIED1.2Funding rounds (last 18 months)SH5 CLAIMSVERIFIED1.3Hiring patterns + role mixQA4 CLAIMSRESEARCHED2.1Product roadmap signalsSH6 CLAIMSRESEARCHED2.2Customer review themesTR3 CLAIMSOPEN3.1Win/loss against usQA3 CLAIMSSYNTHESIZEDCLAIM BOARD · 28 TOTAL · 22 VERIFIEDC-21Competitor X raised $42M Series C, 18 months after Series BSHOWNS4 SOURCESHIGHC-22Pricing moved from per-seat to consumption in Q1; 17% net price increaseQAOWNS3 SOURCESHIGHC-23New competitive product targets enterprise segment with custom pricingTROWNS2 SOURCESMEDIUMC-24Hiring spike in design (4× MoM) suggests product polish pushSHOWNS1 SOURCESUNCONFIRMEDC-25Win-rate against us in regulated verticals dropped 8pp YTDQAOWNS3 SOURCESHIGHC-26Customer reviews flag onboarding friction (mention rate +220%)TROWNS5 SOURCESHIGHCLAIMHIGH CONFIDENCE (≥3 SOURCES)MEDIUM (2 SOURCES)UNCONFIRMED (1 SOURCE)VERIFIED BY QUANTTRIANGULATOR REVIEW
An example research pack's memory browser. Yours can look different — you define which agents own sourcing, triangulation, math, and synthesis for your domain.
What the pack does

Five concrete tasks.

  1. 01

    Specialist agents for every layer of research.

    Compose the pack you need: a Source Hunter for primary sources, a Triangulator for verification, a Quant Analyst for numerical claims, a Synthesizer for the final memo or deck. Add or drop specialists per domain — legal needs a citation specialist, equity research needs filings analysis, market research needs a survey aggregator.

  2. 02

    Methodology loads into every agent.

    A methodology custodian (or your standards memory) holds the citation rules, source-quality thresholds, statistical-verification protocol, and approved sources. Every other agent — Hunter, Triangulator, Quant, Synthesizer — loads it before acting. The same rigor applies whether the question is legal, financial, real-estate, or scientific.

  3. 03

    Triangulate claims, not vibes.

    Triangulator requires multiple independent primary sources for any high-confidence claim. Single-source findings get marked unconfirmed and flagged for review. Hallucinated citations don't pass the verification step because the URL doesn't actually contain the cited claim.

  4. 04

    Crosscheck the math.

    Quant Analyst verifies every statistical claim. Re-runs the math from the raw data when possible — connects to Wolfram, Python, or your data warehouse via MCP. If the math doesn't reproduce, the claim doesn't ship.

  5. 05

    Synthesize without losing fidelity.

    Synthesizer drafts the memo, deck, briefing note, or executive summary. Citations stay attached at the claim level. Knowledge Graph captures entity relationships over time so subsequent research loads structural context.

How a team uses this

Setup once, then watch it run.

Concrete operator setup, the phases the pack moves through, and where you stay in the loop.

Define the research domain — markets you cover, sources you trust, the questions you keep asking.
Connect your archive (Notion, Drive, or a folder) so memos land where you read.
Whitelist primary sources and rank them — the pack weighs them above general web.
Set yourself as the brief-publish approver. The pack never delivers without sign-off.
Phase 1

Question

An analyst (you) drops a research question. Planner sharpens it, breaks it into sub-questions, and assigns each one to a researcher.

Phase 2

Research

Researchers work in parallel, each on its slice. They pull from primary sources first, fall back to general web, and capture every cite in memory.

Phase 3

Synthesize

Synthesizer pulls the slices together, reconciles contradictions explicitly, and surfaces the open questions worth a second pass.

Phase 4

Deliver

You approve and the memo files into your archive with the full citation tree. Next time you ask a related question, the pack starts from the prior memo.

PHASE 01QuestionPHASE 02ResearchPHASE 03SynthesizePHASE 04DeliverApprovedAnalystDROPS QUESTIONPlannerDECOMPOSESQuestion treeSUB-QUESTIONSPrimary sourcesWHITELISTEDResearcher ASLICE 1Researcher BSLICE 2Researcher CSLICE 3SynthesizerMERGES + RECONCILESFact-checkerCITATION PASSMemo draftVERSIONEDApprovalYOU SIGN OFFArchiveNOTION / DRIVEMemory recordDURABLE
An example pack + workflow

Your pack, your workflow.

Workflows are markdown that reference the roles in your pack. Below is one example shape - yours can have different agents, different steps, different cadence.

  1. Step 01Day 0

    Scope

    ReceivesResearch question + deadline
    RLResearch LeadMCMethodology

    Research Lead decomposes the question into 3-7 sub-questions. Each gets a specialist, a deadline, and an output shape (claim, finding, dataset).

    Hands offSub-question list with assignments
  2. Step 02Days 1-3

    Parallel sourcing

    ReceivesSub-questions
    SHHunterSLSource Library

    Source Hunter + assigned specialists work in parallel. Each runs the deep-research skill: ≥3 primary sources per claim, recorded to memory with full citations.

    Hands offSource notes per claim
  3. Step 03Day 4

    Triangulation + verification

    ReceivesSource notes
    TRTriangulatorQAQuant

    Triangulator demands multi-source verification - single-source claims downgrade to unconfirmed. Quant Analyst re-runs every statistical claim from raw data.

    Hands offConfidence-scored findings
  4. Step 04Day 5

    Synthesis

    ReceivesVerified findings
    SYSynthesizerKGKnowledge Graph

    Synthesizer drafts the memo or deck - citations stay attached at claim level. Knowledge Graph captures entity relationships for future projects.

    Hands offDraft document with citations
  5. Step 05Day 5

    Review + publish

    ReceivesDraft document
    RVReviewerRLResearch Lead

    Reviewer runs pre-publish QA against methodology. Research Lead approves before any external distribution.

    Hands offPublished memo + updated graph
Example workflow · 5 steps · per-project handoff chain
Workflow · markdownExample research project lifecycle
# Example research project lifecycle
# Workflows are markdown — yours can swap agents,
# tighten methodology, change the output shape, or
# define a different verification standard.

Match: products labelled "research"
Required tools: web_search, file_read, file_write, http_request
Required skills: deep-research, citation-discipline, statistical-verification

## Methodology context — always-on
Every agent loads from your methodology memory: citation rules,
source-quality thresholds, statistical-verification protocol,
approved sources, do-not-cite list. No agent acts without
methodology context.

## Step 1 — Scope (Research Lead)
Decompose the question into 3-7 sub-questions. Each sub-question
gets a specialist, a deadline, and an output shape (claim,
finding, dataset).

## Step 2 — Parallel sourcing
Source Hunter + assigned specialists work in parallel. Each runs
the deep-research skill: ≥3 primary sources per claim, recorded
to memory with full citations.

## Step 3 — Triangulation
Triangulator reviews findings. Single-source claims downgrade
to unconfirmed. Multi-source claims get a confidence score.

## Step 4 — Numbers verification
Quant Analyst re-runs every statistical claim from raw data.
Mismatches flag the claim for re-verification.

## Step 5 — Synthesis + publish
Synthesizer drafts the memo or deck; Knowledge Graph updates
entity relationships; Reviewer signs off; Research Lead approves
external distribution.

Approvals: Research Lead signs off before any external
distribution.
Built on
CapabilityMemoryCapabilitySkillsCapabilityWorkflowsCapabilityEvals
The ROI

A junior analyst pod's first quarter of work, every week — composable, methodology-loaded, with citations that hold up under scrutiny.

Common questions

Things teams actually ask.

No. Research teams shape per domain — legal teams want a dedicated case-law specialist and a redline agent; equity research teams want a filings analyst and a model-builder; market research teams want a survey aggregator and a competitor watcher; real estate teams want a comp analyst and a deal sourcer. PACKWOLF lets you compose the specialists your domain actually needs. The example here is one shape, not the shape.

Related use cases
Use caseCompliance monitoringContinuous-tracking pattern, regulatory domainUse caseSoftware developmentAdjacent, technical researchUse caseMarketing automationWhere the research surfaces as content

Run this pack on your team's work.

Closed-beta cohorts are small. Tell us about your work and we'll configure the pack for what you actually do.

Request beta access