Agentic Workflows

Your agents run while you sleep.

You set the direction. They do the work.

Get STEEPWORKS

Your AI workforce.

Background research while you shower. Plan and execute while you commute. Finished output waiting when you sit down.

Build personal intelligence feeds

200+ sources scanned and summarized before your alarm.

Create and promote your ideas

Publish like you have a content team across blog, LinkedIn, newsletter, Twitter.

Walk in ready

Company, contact, market context pulled before the call.

Find why deals stall

Surface patterns from CRM, activity, calls, tickets and signals.

Nothing falls through

Research flows to plans. Plans flow to tasks. Your strategy stays current.

The Research Agent

8 agents. One command. Complete prospect brief.

New opportunity enters pipeline → 8 specialized agents fire in sequence.

Phases
Powered By
●

CRM Context

Relationship history, deal stage

HubSpot/SFDC/Attio/Any MCP
●

Company Intelligence

News, Signals, Basics

unified_company_research.md·Perplexity API
●

Tech Stack Discovery

Tooling, complementary tooling

tech_stack_discovery.md·Apify·Firecrawl
●

Pain Signal Mining

Customized to your ICP

pain_intelligence.md·Job posting analysis
●

LinkedIn Discovery

Find and map stakeholders

linkedin_profile_finder.md·WebSearch
●

Profile Extraction

Screenshot, career history, posts

Playwright MCP·Claude Vision
●

Buying Committee

MEDDPICC roles, influence

contact_verification.md·MEDDPICC framework
●

Brief Synthesis

Talk track, objection prep

sales_brief_synthesizer.md

Output: Complete intelligence brief

.md promptsExternal APIsMCP tools

How it actually works.

Trigger. Execute. Customize. Review.

Agent Workflow Coordination
## Research Agent Triggered
 
**Phase 1:** CRM Context
**Phase 2:** Company Intelligence
**Phase 3:** Tech Stack Discovery
 
**Phase 4-6:** *parallel execution*
• Pain signals identified: 4
• LinkedIn profiles found: 7
• Profiles extracted via Vision
 
**Phase 7:** Buying Committee Mapped
**Phase 8:** Brief Synthesized
 
## Complete - Full brief delivered
01STEP

One command fires 8 agents

Spawn specialized agents for each task. Control models, cost, and approaches.

workflow_protocol.md
## Batch Approval Pattern
 
**Problem:** Manual approval for every URL = 10-15 interruptions per agent
 
**Solution:** Batch discovery → single approval → complete extraction
 
1. Each agent uses WebSearch FIRST (autonomous)
2. Agent identifies critical pages needing extraction
3. Agent returns findings + list of critical URLs
4. **You approve batch once**
5. Agent fetches all pages and completes work
 
**Result:** 10-15 interruptions → 1 approval per agent
02STEP

Each step carefully optimized

Each skill has been used hundreds of times. You benefit from the details we've sweated.

research_brief_output.json
{
  "executive_summary": {
    "opportunity": "Strong fit - 3 of 4 pain signals match ICP",
    "urgency": "high",
    "key_insight": "VP Marketing hired 2 months ago, inheriting messy tech stack"
  },
  "talk_track": {
    "opener": "New VPs who own pipeline can't afford tool sprawl slowing them down...",
    "pain_bridge": "New leaders inherit chaos. 6 tools, no single view of pipeline...",
    "value_prop": "What if your CRM actually told you which deals need attention?"
  },
  "next_steps": [
    { "action": "Book discovery call", "owner": "AE", "timeline": "48h" }
  ]
}
03STEP

Get structured and auditable data

Each agent stores its data. This allows you to use it anywhere, identify breakdowns, and iterate quickly.

sales_brief_synthesizer.md
## Phase 5: Sales Brief Synthesis
 
**Agent:** `sales_brief_synthesizer.md` (Opus 4.5)
 
**Synthesizes ALL phases into actionable intelligence:**
 
1. Transform evidence into sales strategy
2. Generate personalized talk track using:
   - Company context
   - Individual's public statements and interests
   - Pain points mapped to value proposition
3. Prepare objection handling based on background
4. Recommend content assets (case studies, posts)
5. Define action items with owners and deadlines
 
**Output:** Executive summary, MEDDPICC assessment,
talk track, objection prep, action items
04STEP

Use (or chat) with the data

Use it to prep for calls. Chain it to other agents. Get alerts when something shifts. Structured data that fits how you actually work.

Equip your agents with dedicated tools

Perplexity API

Company intelligence

HubSpot MCP

CRM context

Apify + Firecrawl

Tech stack discovery

Playwright MCP

LinkedIn extraction

Wire in any tools you have. These are just patterns.

The compound effect.

Day one, they're skilled. Week eight, they're yours.

WEEK 1
  • ✓
    STEEPWORKS-trained

    → Industry best practices

  • ✓
    Production-ready

    → Works out of the box

  • ~
    Minimal context on you

    → Learning starts now

8 weeks
WEEK 8
  • ✓
    Feedback absorbed

    → Knows what you'll cut

  • ✓
    Context loaded

    → References your work

  • ✓
    Voice learned

    → Sounds like you

Architect once. Leverage forever.

Work that does itself.

These aren't demos. They're running in production, right now, doing real work.

Running in Production

Your newsletter writes itself

Every week. On schedule.

Monday morning, a draft lands in your inbox. Topics pulled from your idea bank, written in your voice, formatted and ready. You edit, approve, send. The 8-hour grind? Gone.

Know everything before the call starts

Full brief waiting before you open your calendar.

New deal enters your pipeline. By the time you see it, there's a full brief waiting: company intel, key contacts, talking points, landmines to avoid. Walk in prepared, not scrambling.

Your deals tell you when they need help

Calls + CRM. One intelligence stream.

Objection patterns. Stalled momentum. At-risk deals. Pulled from hundreds of calls and your entire pipeline—synthesized into what actually needs your attention right now.

Never get blindsided by a competitor

Changes detected. Battlecards updated.

When competitors ship, change pricing, or update messaging—you know. Your battlecards stay current without someone manually checking websites every week.

Your contract review happens automatically

Red flags surfaced before signature.

MSAs. NDAs. Employment agreements. Option grants. Each one scanned, risks scored, and questionable clauses flagged. You either sign with confidence—or know exactly what to escalate.

Your system gets sharper every month

Intelligence that compounds.

After 30 days of use, the system has learned your patterns — what you work on, how you work, and what produces the best results. It starts recommending what to focus on next. Not a to-do list. Priorities based on what actually moves the needle.

Your agents work weekends. You don't have to.

Your agents are waiting.

Day one, they run production workflows. Week eight, they think like you. And they never stop improving.

Get STEEPWORKS