Released v0.4.2024

Kalasar MCP Integration

Connect Claude directly to Kalasar.ai's campaign engine. Author sharp brief briefs, update organizations' persistent memories, and launch complete campaigns from a conversational interface.

🎮 Try the Interactive Arcade Tour

Don't want to copy paste config strings? Click through the live simulation of the Kalasar MCP installation right inside your browser!

Launch Tour

1. Connecting the Kalasar MCP Server

The Model Context Protocol (MCP) acts as the secure bridge allowing Claude to run back-end tools on the Kalasar platform. Complete these steps to configure your custom connector.

1
Navigate to Customize
Open the Claude interface and click the "Customize" tab in the left-hand navigation sidebar.
2
Configure Custom Connector
Click the "+" next to Connectors, select "Add custom connector", enter name Kalasar and URL https://hub.kalasar.com/mcp/, then click Add.
3
Grant OAuth Permissions
An authorization browser tab will open for Kalasar.ai. Click Allow Access to authorize Claude to interact with your workspace.
4
Verify Connected Connector
Claude will return to Customize and show "Kalasar (Custom)" connected successfully, registering the 11 integration tool endpoints.
Claude.ai - Customize Settings
Claude home screen with sidebar open showing the Customize tab
Add Custom Connector modal popup with Kalasar URL
Kalasar OAuth permission authorization panel
Successful Kalasar custom connector connection status
Server Address to Copy: Use this hotcode server address when filling Claude's URL prompt.

2. Uploading the Campaign Skill

A custom skill acts as the logic director for Claude. It guides you dynamically through building high-converting briefs and orchestrates calls to the MCP tools we just connected.

1
Open Skills Management
Inside the Claude Customize sidebar, click on the "Skills" tab to view loaded capabilities.
2
Upload kalasar-campaign.zip
Click "+" next to Skills, select "Upload a skill", and drag-and-drop the kalasar-campaign.zip archive file.
3
Trigger Skill Command
Open a fresh Claude chat session, type /kalasar-campaign, and select the skill trigger from the list.
4
Run Automated Campaign
Provide your business facts or sales transcript, and let the agent orchestrate your complete Meta ad campaign assets.
Claude.ai - Customize Settings
Claude skills listing tab view
Drag and drop kalasar-campaign.zip skill archive
Claude chat bar with kalasar-campaign command
Kalasar Hub campaign dashboard view
Important Note on ZIP Uploading: While standard Claude skill uploads restrict compressed folders, Kalasar's campaign skill utilizes an optimized server-unpack configuration mapping multiple references inside the folder. Drag-and-dropping kalasar-campaign.zip is the required method for registering this specific skill suite.

3. Connected Kalasar Tools Reference

When connected, the Kalasar MCP server registers 11 custom tools inside Claude. The assistant can execute these to interact with organization structures, persistent company settings, and live Meta campaign outputs.

READ list_organizations
Usage & Role

Determines which Kalasar workspace context the user is currently operating in. Run this read command before updating styling rules or publishing briefs when multiple accounts exist.

Parameters
No required parameters. Returns array of { org_id, org_name, plan }.
READ get_company_context
Usage & Role

Pulls the persistent organization memory of a designated business entity. Contains durable data including product mechanics, ideal customer profiles, geography filters, and target demographics.

Parameters
org_id (string) - Unique identifier of the organization.
WRITE update_company_context
Usage & Role

Updates persistent business facts. Run whenever the user reveals general, durable corporate changes (e.g., "we no longer sell to residential addresses") rather than transient campaign-specific features.

Parameters
org_id (string) - Unique organization identifier context_payload (object) - New corporate metrics to append
READ get_styleguide
Usage & Role

Fetches corporate tone guides, voice structures, forbidden vocabulary lists, and jargon guidelines. Protects creatives from deploying non-compliant language assets.

Parameters
org_id (string) - Organization target ID.
WRITE update_styleguide
Usage & Role

Appends word constraints and stylistic instructions directly into the corporate style guide. Ensures future generations automatically inherit voice safeguards.

Parameters
org_id (string) - Organization target ID style_payload (object) - New voice rules, jargon exceptions, or banned words
READ list_campaigns
Usage & Role

Queries historical meta ad runs on the Kalasar platform. Prevents pitch duplication and helps Claude identify historical, high-performing creative hypotheses.

Parameters
org_id (string) - Target workspace.
WRITE create_campaign_generation
Usage & Role

The core transactional operation. Invokes the back-end ad-rendering pipeline once a brief has been completed and verified by the customer. Do not execute speculatively on a draft.

Parameters
org_id (string) - Organization target ID brief_markdown (string) - Fully structured brief following template parameters

4. The Campaign Brief Blueprint

The brief is Kalasar's primary artifact for capturing what makes a campaign sharp. It is not an intake form to copy-paste to customers—it is written based on high-touch discovery sessions, pre-filled with data contexts, or iterated dynamically.

The Three-Phase Relationship Lifecycle:
  • Phase 1 (Onboarding): Kalasar interviews the customer (60-90m call) and writes the brief behind the scenes. Never send a blank form to a new customer.
  • Phase 2 (Steady State): An LLM pre-fills the brief from sales logs, transcripts, and Slack context. The customer reviews and tightens the draft.
  • Phase 3 (Self-Serve): Established customers author briefs directly for minor frame iterations, accelerating campaign launches.
Section 1 - Target Configuration

Best for established offers with precise known buyers, regional services, or tight B2B ICPs. Meta will optimize strictly within a defined structural audience.

Section 1.1A Example Profile "Owners of engineering and architecture firms with 5-20 employees, who personally make software purchasing decisions and upload massive files daily."

Best for new product launches, positioning experiments, or untested markets. No hard audience boundaries are set—instead, you deploy a highly polarizing problem statement, and let Meta route to segment buyers who lean in.

Section 1.1B Example Problem "You are paying for an outdated enterprise ERP, and you need a consultant-developer just to customize a single invoice template. It is 2026."
Section 3.2 - Choose Psychological Frames

One brief can support multiple ad hooks, but each hook must commit to one psychological framing. Never blend framings in a single ad variant. Contrast them to test different buyer motivations.

Latent Loss Curiosity
"Something is silently costing you. Here is how it is leaking."
Dated Urgency Deadline
"Act before [Date] or pay the contract renewal penalty."
Active Fire Crisis
"Your system is bleeding money right now. Stop the leak."
No-Brainer Asymmetric
"Frictionless takeover with guaranteed zero double-billing."
Status/Peer Status
"Your peer agencies have already switched. Don't fall behind."
Before/After Contrast
"Here is your mess. Here is our setup. Cross the short bridge."

Latent Loss / Discovery Frame

Curiosity & Mild Self-Recognition

Core Concept: Highlight a silent, bleeding cost that the buyer is currently unaware of. Guide them toward self-diagnosing the issue.

Ideal Buyer Profile

Discovery-curious, analytical, value-conscious buyers who want to ensure they are not overpaying or falling behind.

Verbatim Listening Guide

Listen for: "I didn't even know we were paying that," or "It felt like a leak we couldn't locate."

Copy Campaign Brief Template (.md)
Markdown template 
# Campaign Brief

This brief tells our pipeline what to build. Vague answers produce vague ads.

## Section 1 — Who is this for?
Your chosen mode (A - Capture or B - Discovery): [A or B]

### If Mode A (Capture):
1.1A Who is the sharpest version of your buyer? (Industry, size, role)
> Your answer: 

1.2A How tightly can we target geographically while still hitting enough buyers?
> Your answer: 

### If Mode B (Discovery):
1.1B What problem statement, said plainly, makes the right buyer stop scrolling?
> Your answer: 

1.2B Who do you suspect might lean into this problem? List 2–4 buyer types.
> Your answer: 

1.3B Geofilters:
> Your answer: 

---

## Section 2 — What counts as success?
2.1 What does a qualified lead look like for you? (Operational criteria)
> Your answer: 

2.2 What does a junk lead look like? (Clear negative patterns to exclude)
> Your answer: 

---

## Section 3 — The bet
3.1 What is the one realization we want to put in the buyer's head?
> Your answer: 

3.2 Which frames will we use to land that realization? (Select Primary + 1-2 contrast variants)
> Primary Frame: 
> Variant 2 Frame: 

3.3 What is the cost of inaction we want the buyer to feel? (Time, money, productivity leaks)
> Your answer: 

3.4 What is the one reason the buyer should pick you that alternative setups can't claim?
> Your answer: 

---

## Section 4 — The buyer in their own words
4.1 What does your buyer's actual language sound like? Three real sentences in their voice.
> "..."
> "..."
> "..."

4.2 Walk us through the last buyer who said yes — what was the moment they decided?
> Your answer: 

---

## Section 5 — The offer
5.1 What does the buyer get, in the words they'd use telling a colleague?
> Your answer: 

5.2 What is the felt result, in one plain sentence the buyer would say?
> Your answer: 

5.3 How can we create a frictionless journey? (Estimated steps with times)
> Your answer: 

5.4 How can we lower the perceived risk of the buyer? (Guarantees, takeovers, etc. If none, leave blank)
> Your answer: 

5.5 What language traps could break trust or sound off to your buyer? (Acronyms, competitor vocabulary)
> Your answer: