Getting Started
1. Open Studio and describe your goal
Go to Studio and type what you want to optimise: "Audit my campaigns for wasted spend", "Monitor competitor prices daily", or click a domain chip. No account connection needed — Decisify will build an agent, generate a plan, and propose actions right away.
2. Connect a data source (optional)
For live data, go to Connections and link a platform (Google Ads, Shopify, etc.), upload a CSV, or paste a URL. Decisify syncs and indexes your data automatically. You can also scrape any public URL directly from Studio without connecting an account.
3. Review and apply
Every proposed change shows reasoning and evidence. Approve what looks right, reject what doesn't. When ready, hit Apply — changes go live on the connected platform. Every applied run has a one-click Rollback.
Install the CLI
npm install -g @decisify/cli
decisify login
decisify run --agent <agent_id>Quick API call
curl -X POST https://api.decisify.io/api/v1/runs \
-H "Authorization: Bearer <your_token>" \
-H "Content-Type: application/json" \
-d '{"agent_id": "<id>", "account_id": "<account_id>"}'Core Concepts
Decisify is built around a few key ideas that make autonomous agents safe, transparent, and controllable — across any domain.
Shadow mode (default)
In shadow mode, agents read your data but never write. All runs produce plans you can review. Nothing changes until you click Apply.
Live mode
In live mode, approved changes are applied directly to the connected platform after you click Apply. You can still review every change before applying.
Action plans
Every run produces an ActionPlan — a structured list of proposed changes with evidence and reasoning. You review the plan, approve or reject individual actions, then apply.
Connections
Connections link Decisify to your platforms and data sources — via OAuth, file upload, or URL. Google Ads has full read/write support. Shopify, CSV, and URL scraping are live. More platforms are on the roadmap.
Agent Pathways
Decisify offers 4 pathways for building agents — from simple rules to full custom code. All pathways are available on every plan.
Rules & Policies (Pathway A)
Build IF/THEN guardrails: pause keywords above CPA, cap bid changes, exclude campaigns, set alerts. Rules become hard guardrails on every run.
Config-as-Code (DSL/YAML) (Pathway B)
Define strategies as versioned YAML configs — scoring weights, action allowlists, thresholds. Testable, reviewable, diffable.
ML / Adaptive (Pathway C)
Self-improving agents that learn from data. Bandits, forecasting, anomaly detection, hypothesis testing, confidence intervals. ML proposes, guardrails decide.
Custom Code (Sandbox) (Pathway D)
Write agent logic in Python/JS. Your code outputs an ActionPlan — Decisify handles execution, safety, and all mutations with full audit and rollback.
Example YAML strategy (Pathway B)
strategy: pause_high_cpa_keywords
thresholds:
cpa_multiplier: 1.5
min_conversions: 10
actions:
- type: pause_keyword
condition: cpa > threshold.cpa_multiplier * account_avg_cpa
guardrails:
- max_pauses_per_run: 20
- exclude_campaigns: [branded]Runs & Actions
A run is a single execution of an agent against your connected data. Every run follows a consistent lifecycle.
Run lifecycle
Actions
Each run produces a set of actions — individual changes like pausing a keyword, adjusting a bid, or updating a budget. Every action includes the reasoning, evidence metrics, and expected impact.
Rollback
Every applied run has a one-click Rollback that reverts all changes. The audit log records exactly what changed and when, so you always have a full history.
Starting a run via API
curl -X POST https://api.decisify.io/api/v1/runs \
-H "Authorization: Bearer <your_token>" \
-H "Content-Type: application/json" \
-d '{"agent_id": "agt_abc123", "account_id": "acc_xyz789"}'Policy Engine
The policy engine is Decisify's safety layer. Every proposed action passes through policy checks before being shown to you. Actions that exceed your configured risk thresholds are blocked automatically.
Default policies
- *Maximum budget increase per run: 20%
- *Maximum bid adjustment: +/- 30%
- *Minimum data threshold before destructive actions
- *Branded/critical entity protection
- *Rate limiting: max changes per run
Custom policies
You can define custom policies in YAML or through the UI. Policies are versioned and auditable. They apply across any domain — ads, pricing, inventory, or custom agents.
policy: conservative_bidding
rules:
- action: adjust_bid
max_change_pct: 15
require_min_clicks: 50
- action: pause_keyword
require_min_conversions: 20
exclude_match_types: [exact]Safety & Governance
Decisify is designed so that autonomous agents stay under your control. Every layer — from shadow mode to audit logs — exists to make sure nothing changes without your knowledge and consent.
Shadow mode
Every new agent starts in shadow mode. The agent reads your data, analyzes it, and proposes actions — but never writes anything. You review the full action plan before anything goes live. Switch to live mode only when you're confident.
Approval workflows
Each run produces an action plan that requires explicit approval. You can approve all actions, reject individual ones, or send the plan back for revision. On Team and Enterprise plans, multi-person approval chains are available — require a second reviewer before high-impact changes go live.
Spending & rate caps
Set hard limits on what agents can do per run: maximum budget increase, maximum bid adjustment, maximum number of changes, and minimum data thresholds before destructive actions. Caps are enforced by the policy engine and cannot be overridden by the agent.
Audit logs
Every action — proposed, approved, rejected, applied, or rolled back — is recorded in an immutable audit log. See who did what, when, and why. Filter by agent, user, date, or action type. Export logs for compliance reporting.
Rollback
Every applied run has a one-click Rollback that reverts all changes made during that run. Rollbacks are themselves audited, so you always have a complete record. If an agent makes a mistake, you can undo it in seconds.
Role-based access
Control who can create agents, approve runs, apply changes, and manage connections. Admins configure policies; operators run agents; viewers see results. Available on Team and Enterprise plans.
Build Anything
Decisify isn't just for ads. Any goal that involves analyzing data and proposing actions can become an agent. Here are 10 real examples — paste any of them into Studio to get started.
"Monitor competitor prices on Amazon for my top 50 SKUs and flag when I'm more than 10% above the lowest price"
Price monitoring agent that scrapes competitor listings and recommends price adjustments.
"Audit my Google Ads account for wasted spend — find keywords with high cost and zero conversions in the last 30 days"
Waste detection agent that identifies underperforming keywords and recommends pauses.
"Detect stockout risk across my Shopify catalog — flag products where current stock is below 2 weeks of sales velocity"
Stockout prediction agent using sales velocity and current inventory levels.
"Audit monthly vendor invoices for anomalies — flag any invoice that's more than 2 standard deviations from the 6-month average for that vendor"
Spend anomaly detection agent that catches unusual invoices before they're paid.
"Identify my top 20 customers by lifetime value and find which acquisition channels they came from"
Customer LTV attribution agent for optimizing acquisition spend.
"Scrape the top 10 Google results for 'best CRM software 2026' and summarize the key features each competitor highlights"
Competitive research agent that gathers and synthesizes publicly available information.
"Build an agent that tracks open headcount across departments and flags roles that have been open for more than 45 days"
Recruiting pipeline monitor for workforce planning.
"Recommend optimal pricing tiers for my SaaS product based on competitor pricing data and my current conversion rates"
Pricing strategy agent that combines market data with your own metrics.
"Collect and summarize the latest 20 papers on transformer efficiency from arXiv, grouped by technique (pruning, quantization, distillation)"
Academic research agent for staying current on a topic.
"Scan my product descriptions for GDPR-sensitive language and flag any that mention personal data collection without a privacy link"
Compliance scanning agent for regulatory risk detection.
API Reference
The Decisify API is a REST API. All endpoints require a Bearer token.
Base URL
Authentication
Authorization: Bearer <your_token>Get your token from Settings → API.
Endpoints
Example: Create an agent
{
"name": "CPA Guardian",
"type": "yaml_strategy",
"account_id": "acc_xyz789",
"config": {
"strategy": "pause_high_cpa",
"thresholds": {
"cpa_multiplier": 1.5
}
}
}All endpoints return JSON. Authentication via Bearer <token> header. Base URL: https://api.decisify.ai
FAQ
Frequently asked questions about Decisify.
What is Decisify?
Decisify is an agentic optimisation platform. You describe a goal — cut ad waste, monitor competitor prices, detect stockouts, audit spend — and Decisify builds an autonomous agent that plans, executes, and improves. Every recommendation comes with evidence and reasoning, and nothing changes without your approval.
Is my data safe?
Yes. Decisify uses OAuth 2.0 for platform connections — we never store your passwords. Your data is encrypted in transit (TLS 1.3) and at rest (AES-256). We run in SOC 2 compliant infrastructure, and you can revoke access at any time from your account settings. We never share your data with third parties.
What platforms and domains are supported?
Decisify works across ads, e-commerce, pricing, ops, growth, and research. Google Ads has full read/write support including MCC accounts. Shopify, CSV/file upload, and URL scraping are supported. Meta Ads, GA4, HubSpot, Stripe, BigQuery and more are on the roadmap.
Can I undo changes?
Absolutely. Every applied run includes a one-click Rollback button that reverts all changes made during that run. The full audit log records exactly what was changed, when, and why — so you always have complete visibility and control.
What's the pricing?
We offer a generous free tier for individuals and small teams. For pricing details, plans, and enterprise options, visit our pricing page.