AI Agent Platform

How It Works

The Agent Platform gives AI tools a secure way to act on your behalf. You create a scoped API token, hand it to any AI tool, and the tool can then call actions like "create a QR code" or "draft an email blast" through a standard REST API.

Scoped tokens — each token only has access to the scopes you grant. An SMS token can't touch your BuzzPins.
Human approval queue — high-risk and outreach actions (email, SMS, payments) are held until you click Approve.
Immutable audit log — every action run is recorded forever with its input, output, and approval record.
MCP + OpenAPI — compatible with Claude projects, ChatGPT GPTs, and any tool that reads OpenAPI specs.

No coding required. If you're using ChatGPT or Claude, you paste one URL and the AI discovers all actions automatically. Setup takes under 5 minutes.

Quick Setup (5 Minutes)

1

Log in and open your Agent Dashboard

Go to /admin/agents. This is your control center — token management, action history, and approval queue all live here.

2

Create your first agent token

In the dashboard, click Create Token. Give it a name (e.g. "ChatGPT Assistant"), choose the scopes you want, and optionally set an expiry. You'll see your token once — copy it immediately. A copy is also emailed to you.

3

Or create a token via API

If you prefer the API, POST to /api/v1/agent/token while logged in:

# Create a token (must be logged in to the web app first)
curl -X POST https://itsbuzzing.com/api/v1/agent/token \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My ChatGPT Assistant",
    "client_type": "chatgpt",
    "scopes": ["buzzcards.write", "content.generate", "referrals.write"],
    "expires_in_days": 90
  }'

# Response — copy the token now, it won't be shown again
{
  "token": "bza_your_raw_token_here",
  "prefix": "bza_a1b2c3d4e5f6g7h8...",
  "client_id": "uuid",
  "scopes": ["buzzcards.write", "content.generate", "referrals.write"],
  "note": "Store this token securely — it is shown only once."
}

4

Test your token

Run a dry-run to confirm the token works and see what the approval policy would be, without executing anything:

curl -X POST "https://itsbuzzing.com/api/v1/agent/actions/generate_qr_code/run?dry_run=true" \
  -H "Authorization: Bearer bza_your_token_here" \
  -H "Content-Type: application/json" \
  -d '{"target_url": "https://example.com", "label": "My QR"}'

# Returns policy info — no action executed
{
  "dry_run": true,
  "policy": { "requires_human_approval": false, "risk_level": "low" },
  "message": "Dry run only. No action executed."
}

Connect ChatGPT (Custom GPT)

1

Create a token with chatgpt type

Use "client_type": "chatgpt" when creating the token. Grant the scopes you want ChatGPT to access.

2

Open your GPT editor and add an Action

In ChatGPT, go to Explore GPTs → Create → Configure → Actions → Add Action.

3

Import the OpenAPI spec

Click Import from URL and paste:
https://itsbuzzing.com/api/v1/mcp/openapi.json
ChatGPT will auto-discover all available actions.

4

Set authentication

Choose API Key → Bearer and paste your token (bza_...). Save.

Approval-required actions: ChatGPT will submit the action and receive a 202 awaiting_approval response. You'll get an email to approve it from your dashboard before anything executes.

Connect Claude (MCP Project)

1

Create a token with claude type

Use "client_type": "claude" and grant the scopes you want Claude to use.

2

Add the MCP manifest to your Claude project

In your Claude project settings, add a custom tool source and paste:
https://itsbuzzing.com/api/v1/mcp/manifest.json
Claude will read all action definitions automatically.

3

Set the bearer token in project instructions

In your project instructions, tell Claude: "Always include the header Authorization: Bearer bza_your_token when calling It's Buzzing actions."

Claude reads the manifest's requiresApproval field and will warn you proactively before submitting high-risk actions.

Connect Zapier (No-Code Automations)

1

Create a token with custom type

Use "client_type": "custom" and scope it to just what your Zap needs (e.g. buzzcards.write).

2

In Zapier, add a Webhooks by Zapier step

Choose POST, set the URL to the action you want, e.g.:
https://itsbuzzing.com/api/v1/agent/actions/create_buzzcard/run

3

Add the auth header

Under Headers, add:
Authorization → Bearer bza_your_token_here

4

Map your Zap data to the action payload

Set the Data field to a JSON object with the action's input fields. Example for create_buzzcard:

{
  "business_name": "{{business_name from previous step}}",
  "slug": "{{slug}}",
  "email": "{{email}}",
  "phone": "{{phone}}"
}

Use the Idempotency-Key header in Zapier to prevent duplicate actions if a Zap retries: add header Idempotency-Key → {{zap_id}}-{{step_id}}.

Connect Google Gemini

Gemini uses a Tool Calling interface. You define It's Buzzing actions as tools in your API call, and Gemini decides when to invoke them based on the conversation.

1

Create a token with gemini type

Go to /admin/agents → Create Token. Set "client_type": "custom" (Gemini doesn't have a dedicated type). Recommended scopes: buzzcards.write, content.generate, referrals.write, crm.read.

2

Fetch the MCP manifest to auto-load tools

The manifest at /api/v1/mcp/manifest.json returns all action definitions. Pull them at startup so you don't hardcode schemas:

# Auto-load all It's Buzzing tools from the manifest
import requests

manifest = requests.get("https://itsbuzzing.com/api/v1/mcp/manifest.json").json()

# Convert to Gemini tool format
tools = []
for t in manifest["tools"]:
    tools.append({
        "name": t["name"],
        "description": t["description"],
        "input_schema": t["inputSchema"],
    })

3

Wire up the tool executor and start a chat

import google.generativeai as genai

genai.configure(api_key=YOUR_GEMINI_API_KEY)
ITSBUZZING_TOKEN = "bza_your_token_here"

model = genai.GenerativeModel(model_name="gemini-2.0-flash", tools=tools)
chat  = model.start_chat()

response = chat.send_message("Create a BuzzCard for my salon")

# Handle tool calls
if response.tool_calls:
    for call in response.tool_calls:
        result = requests.post(
            f"https://itsbuzzing.com/api/v1/agent/actions/{call.name}/run",
            headers={"Authorization": f"Bearer {ITSBUZZING_TOKEN}"},
            json=call.args,
        ).json()
        # Send result back to Gemini to continue the conversation
        response = chat.send_message(str(result))

Approval-required actions: Gemini will receive a 202 awaiting_approval response. Have your tool executor return "status": "approval_required — check /admin/agents" so Gemini can relay that to the user.

4

Use idempotency keys to prevent duplicate executions

Gemini may retry tool calls. Add an Idempotency-Key header derived from the conversation turn to deduplicate:

import uuid

idempotency_key = str(uuid.uuid4())  # generate once per turn

result = requests.post(
    f"https://itsbuzzing.com/api/v1/agent/actions/{call.name}/run",
    headers={
        "Authorization": f"Bearer {ITSBUZZING_TOKEN}",
        "Idempotency-Key": idempotency_key,
    },
    json=call.args,
).json()

Connect Perplexity

Perplexity uses the OpenAI-compatible chat completions API with tool calling. It's Buzzing's OpenAPI spec maps directly to Perplexity's tool format — load it once and Perplexity can call all available actions.

What makes Perplexity unique: It combines real-time web search with tool calling. Your assistant can look up current information about a business and take action in It's Buzzing in the same conversation turn.

1

Create a token

Go to /admin/agents → Create Token. Name it Perplexity – Research & Automation. Recommended scopes: all read scopes + content.generate, referrals.write, buzzcards.write.

2

Auto-load tools from the MCP manifest

Perplexity uses the OpenAI function-calling format. Convert the manifest once at startup:

import requests

manifest = requests.get("https://itsbuzzing.com/api/v1/mcp/manifest.json").json()

# Convert MCP manifest → OpenAI/Perplexity tool format
tools = [
    {
        "type": "function",
        "function": {
            "name": t["name"],
            "description": t["description"],
            "parameters": t["inputSchema"],
        }
    }
    for t in manifest["tools"]
]

3

Call Perplexity with tools enabled

Use Perplexity's OpenAI-compatible endpoint with your Perplexity API key:

from openai import OpenAI

ITSBUZZING_TOKEN  = "bza_your_token_here"
PERPLEXITY_API_KEY = "pplx_your_key_here"

client = OpenAI(
    api_key=PERPLEXITY_API_KEY,
    base_url="https://api.perplexity.ai",
)

messages = [{"role": "user", "content": "Set up a BuzzCard for my barbershop"}]

response = client.chat.completions.create(
    model="llama-3.1-sonar-large-128k-online",
    messages=messages,
    tools=tools,
)

# Handle tool calls in a loop
import json

while response.choices[0].finish_reason == "tool_calls":
    tool_calls = response.choices[0].message.tool_calls
    messages.append(response.choices[0].message)

    for call in tool_calls:
        result = requests.post(
            f"https://itsbuzzing.com/api/v1/agent/actions/{call.function.name}/run",
            headers={
                "Authorization": f"Bearer {ITSBUZZING_TOKEN}",
                "Idempotency-Key": call.id,  # Perplexity provides stable call IDs
            },
            json=json.loads(call.function.arguments),
        ).json()

        messages.append({
            "role": "tool",
            "tool_call_id": call.id,
            "content": json.dumps(result),
        })

    response = client.chat.completions.create(
        model="llama-3.1-sonar-large-128k-online",
        messages=messages,
        tools=tools,
    )

print(response.choices[0].message.content)

Tip: Perplexity's tool_call_id is stable per call, so you can use it directly as the Idempotency-Key — no need to generate your own.

4

Handle approval-required actions

When an action returns 202 awaiting_approval, pass the message back to Perplexity so it can tell the user what to do:

# In your tool executor
if result.get("status") == "awaiting_approval":
    content = (
        f"Action submitted for approval. "
        f"Approve it at https://itsbuzzing.com/admin/agents "
        f"(approval ID: {result['approval_id']})"
    )
else:
    content = json.dumps(result)

messages.append({
    "role": "tool",
    "tool_call_id": call.id,
    "content": content,
})

Available Scopes

Every token is granted only the scopes you specify at creation time. You can also check the full scope list at /api/v1/agent/scopes.

buzzcards.readRead BuzzCard profiles

buzzcards.writeCreate & update BuzzCards

buzzpins.readRead BuzzPins listings

buzzpins.writeCreate & update pins

buzzpresence.readRead check-in data

buzzpresence.writeCreate check-in tokens

churchhub.readRead church hub data

churchhub.writeCreate church campaigns

commerce.readRead product/store data

commerce.writeCreate share pages

crm.readRead leads & contacts

crm.writeUpdate CRM records

referrals.readRead referral data

referrals.writeCreate referral links

content.generateAI content generation

email.prepareDraft email campaigns

email.send.requires_approvalSend emails (approval required)

sms.prepareDraft SMS campaigns

sms.send.requires_approvalSend SMS (approval required)

payments.readRead payout/billing data

payments.write.requires_approvalBilling actions (approval required)

public_publish.requires_approvalPublish public content (approval required)

agent.manageManage tokens & governance actions

All Available Actions

Fetch the live list any time at /api/v1/agent/actions. Each action links to its schema. High risk actions always require human approval before execution.

Growth & Presence

Action Key	Risk	Approval?	Description
create_buzzcard	low	—	Create a digital business card with QR code
generate_qr_code	low	—	Create a branded QR code for any URL
create_referral_link	low	—	Generate a trackable referral URL
create_buzzpin	medium	Yes	Drop a location pin on the BuzzPins map
setup_smart_checkin	low	—	Generate an attendance token and check-in URL
create_product_share_page	low	—	Build a shareable product micro-page
create_buzzbattle_matchup	low	—	Launch a community vote bracket
create_creator_daily_question	low	—	Scaffold a BuzzPulse poll for creators

Outreach & CRM

Action Key	Risk	Approval?	Description
prepare_email_blast	high	Yes	Draft an email campaign (held for review)
prepare_sms_blast	high	Yes	Draft an SMS campaign (held for review)
list_leads	low	—	Paginated CRM lead list with filters
update_lead_status	medium	Yes	Move a lead through the pipeline
create_church_engagement_campaign	medium	Yes	Set up visitor follow-up for churches
launch_ai_workspace_plan	high	Yes	Execute a multi-step AI-generated setup plan

Onboarding & Ambassadors

Action Key	Risk	Approval?	Description
signup_user	high	Yes	Create a new It's Buzzing user account
enroll_ambassador	high	Yes	Create ambassador profile with referral code
assign_ambassador_to_campaign	medium	Yes	Link an ambassador to a campaign
payout_preview	low	—	Read-only commission totals (no money movement)
create_or_update_local_listing	medium	Yes	Upsert a BuzzPresence local listing
submit_business_claim	high	Yes	Initiate a business ownership claim

Analytics

Action Key	Risk	Approval?	Description
analytics_overview	low	—	Views, clicks, conversions, and revenue totals
analytics_timeseries	low	—	Daily/weekly metric series for charts
analytics_funnel	low	—	Conversion funnel with drop-off rates
analytics_attribution	low	—	Lead breakdown by source, channel, ambassador
platform_health_summary	low	—	System-wide action success rates and webhook health
action_failure_hotspots	low	—	Top failing actions with error summaries

Webhooks

Action Key	Risk	Approval?	Description
create_webhook	low	—	Register a URL for HMAC-signed event payloads
list_webhooks	low	—	List all registered webhook endpoints
rotate_webhook_secret	medium	Yes	Generate a new HMAC signing secret
replay_webhook_event	medium	Yes	Re-fire a past payload to an endpoint

Governance & Control

Action Key	Risk	Approval?	Description
get_action_run	low	—	Fetch full detail for any action run
audit_log_query	low	—	Query immutable run log with filters
list_action_policies	low	—	Every action with risk level and scope requirements
validate_action_payload	low	—	Schema preflight without executing
simulate_action_policy	low	—	Preview approval policy for an action + payload
toggle_action_enabled	high	Yes	Enable or disable any action in the registry
retry_action_run	high	Yes	Re-run a failed action with the same payload
replay_action_run	high	Yes	Fresh run from a completed action's payload

Human Approval Queue

When an agent submits a high-risk action, execution is paused and you receive an email with Approve / Reject links. You can also manage approvals directly from your dashboard.

What triggers approval?

The action's risk_level is high
The action key starts with a sensitive prefix: signup_, enroll_, submit_, assign_, update_, rotate_, or replay_
The action has requires_human_approval: true in its definition

Approval flow

Agent calls POST /api/v1/agent/actions/<key>/run
Platform returns 202 awaiting_approval with an approval_id
You receive an email with an Approve and a Reject button
Clicking Approve executes the action and fires your webhooks with action.completed
Clicking Reject marks the run as failed and fires action.rejected

Approval requests expire after 48 hours. If you don't act within that window the run stays in awaiting_approval and must be retried.

Managing approvals via API

# List pending approvals
curl https://itsbuzzing.com/api/v1/agent/approvals \
  -H "Authorization: Bearer bza_your_token"

# Approve an action
curl -X POST https://itsbuzzing.com/api/v1/agent/approvals/{approval_id}/approve \
  -H "Authorization: Bearer bza_your_token"

# Reject an action
curl -X POST https://itsbuzzing.com/api/v1/agent/approvals/{approval_id}/reject \
  -H "Authorization: Bearer bza_your_token" \
  -d '{"reason": "Wrong recipient list"}'

Webhooks

Register a URL and receive HMAC-signed POST payloads whenever actions complete, fail, need approval, or are rejected.

Register a webhook

curl -X POST https://itsbuzzing.com/api/v1/agent/webhooks \
  -H "Authorization: Bearer bza_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://your-server.com/hooks/buzzing",
    "event_types": ["action.completed", "action.rejected", "action.failed"]
  }'

Verify the signature

Every delivery includes an X-Buzzing-Signature header. Verify it to reject spoofed requests:

# Python example
import hmac, hashlib

def verify(body_bytes, signature_header, secret):
    expected = hmac.new(secret.encode(), body_bytes, hashlib.sha256).hexdigest()
    return hmac.compare_digest(f"sha256={expected}", signature_header)

Event types

Event	When fired
action.completed	Action executed successfully (including post-approval)
action.failed	Action execution threw an error
action.rejected	Approval was denied by the account owner
webhook.test	Manual test ping from dashboard or API

All Links at a Glance

Integration setup guides (this page)

ChatGPT

GPT Builder + OpenAPI import

#connect-chatgpt

Claude

MCP manifest + project instructions

#connect-claude

Google Gemini

Tool calling + manifest auto-load + idempotency

#connect-gemini

Perplexity

OpenAI-compatible tool calling + approval handling

#connect-perplexity

Zapier

Webhooks step + idempotency key

#connect-zapier

User-facing pages

Marketing / Overview

What the platform is and how it works — share this with teams evaluating it

/agent-platform

Agent Dashboard

Your control center — tokens, run history, approval queue

/admin/agents

This Help Page

Setup guide and API reference

/help/agent-platform

Discovery endpoints (public, no auth)

MCP Manifest

Paste into Claude projects or any MCP-compatible tool

/api/v1/mcp/manifest.json

OpenAPI 3.0 Spec

Import into ChatGPT GPT Actions or Postman

/api/v1/mcp/openapi.json

Action Registry

Live list of all enabled actions with schemas (auth required)

/api/v1/agent/actions

Available Scopes

Full list of valid scope strings for token creation

/api/v1/agent/scopes

Core API endpoints (auth required)

Method	Path	What it does
POST	/api/v1/agent/token	Create a scoped agent token
GET	/api/v1/agent/clients	List all your tokens
DELETE	/api/v1/agent/clients/<id>	Revoke a token permanently
PATCH	/api/v1/agent/clients/<id>/suspend	Temporarily suspend a token
POST	/api/v1/agent/clients/<id>/rotate	Revoke & reissue with same scopes
POST	/api/v1/agent/actions/<key>/run	Execute any action
GET	/api/v1/agent/runs	Action run history
GET	/api/v1/agent/runs/<id>	Single run detail
GET	/api/v1/agent/approvals	Pending approval queue
POST	/api/v1/agent/approvals/<id>/approve	Approve and execute
POST	/api/v1/agent/approvals/<id>/reject	Reject an action
GET	/api/v1/agent/analytics	Action usage stats
GET	/api/v1/agent/analytics/timeseries	Daily/weekly run timeseries
GET	/api/v1/agent/assets	All assets created by agents
GET	/api/v1/agent/webhooks	List registered webhooks
POST	/api/v1/agent/webhooks	Register a webhook
DELETE	/api/v1/agent/webhooks/<id>	Delete a webhook
POST	/api/v1/agent/webhooks/<id>/test	Send a test ping
POST	/api/v1/workspace/generate	Turn a prompt into a setup plan
GET	/api/v1/workspace/<id>	Get workspace project detail
POST	/api/v1/workspace/<id>/execute	Execute approved plan steps

In This Guide

How It Works

Quick Setup (5 Minutes)

Log in and open your Agent Dashboard

Create your first agent token

Or create a token via API

Test your token

Connect ChatGPT (Custom GPT)

Create a token with chatgpt type

Open your GPT editor and add an Action

Import the OpenAPI spec

Set authentication

Connect Claude (MCP Project)

Create a token with claude type

Add the MCP manifest to your Claude project

Set the bearer token in project instructions

Connect Zapier (No-Code Automations)

Create a token with custom type

In Zapier, add a Webhooks by Zapier step

Add the auth header

Map your Zap data to the action payload

Connect Google Gemini

Create a token with gemini type

Fetch the MCP manifest to auto-load tools

Wire up the tool executor and start a chat

Use idempotency keys to prevent duplicate executions

Connect Perplexity

Create a token

Auto-load tools from the MCP manifest

Call Perplexity with tools enabled

Handle approval-required actions

Available Scopes

All Available Actions

Growth & Presence

Outreach & CRM

Onboarding & Ambassadors

Analytics

Webhooks

Governance & Control

Human Approval Queue

What triggers approval?

Approval flow

Managing approvals via API

Webhooks

Register a webhook

Verify the signature

Event types

All Links at a Glance

Integration setup guides (this page)

ChatGPT

Claude

Google Gemini

Perplexity

Zapier

User-facing pages

Marketing / Overview

Agent Dashboard

This Help Page

Discovery endpoints (public, no auth)

MCP Manifest

OpenAPI 3.0 Spec

Action Registry

Available Scopes

Core API endpoints (auth required)