Zapier Agents

Enterprise

Zapier

Product overview

Name of Agent: Zapier Agents
Short description of agent: "platform where you can create custom agents in minutes by describing what you want them to do in plain language. Unlike chat-based agents that only respond when prompted, Zapier Agents proactively monitor for triggers and take action automatically" (link, archived)
Date of release: March 2024 (link) (launched as "Zapier Central", non agentic offering) January 2025 (link) (rebranded to "Zapier Agents") December 2025 (general availability release) [company source]
Advertised use: Automate AI workflows across sales (lead qualification, CRM updates), customer support (ticket triage, response drafting), marketing (content creation, campaign monitoring), operations (meeting prep, data sync), and HR (onboarding automation, hiring (link, archived)
Monetisation/Usage price: free, (link, archived) 50, higher rate limits enterprise custom (link, archived)
Who is using it?: enterprises
Category: Enterprise

Company & accountability

Developer: Zapier
Name of legal entity: Zapier, Inc.
Place of legal incorporation: Delaware (link, archived)
For profit company?: Yes
Parent company?: Zapier
Governance documents analysis: Terms of Service, Privacy Policy, Acceptable Use Policy, and AI-specific policies (link, archived)
AI safety/trust framework: None found
Compliance with existing standards: Zapier have obtained independent third-party auditor certifications with the AICPA’s SOC for Service Organizations, SOC 2 Type II, and SOC 3. (link, archived)also trust.zapier.com (link)

Technical capabilities & system architecture

Model specifications: Zapier Agents uses state of the art models from a variety of model families including OpenAI, Anthropic, Gemini, and open source models (link, archived)
Observation space: User input, files, and applications (link, archived)
Action space: Connectors to many different third party apps, including creating/updating records, sending messages, managing files, triggering workflows, and returning information to users. ((link, archived), company information)
Memory architecture: Agents can be given access to specific data sources (link, archived), but no mention of a memory system beyond this. "Agents do not have persistent memory between session" [company source]
User interface and interaction design: Form-like interface for configuring agent, adding tools, and connecting knowledge bases. Once configured, agents run autonomously in the background. [company source]
User roles: Designer (user is designing an agent to use within Zapier but can interact with external environments and tools)
Component accessibility: Closed source

Autonomy & control

Autonomy level and planning depth: L1-3: user can design an agent without copilot assistance, but copilot can also help modify the agent itself. Copilot asks plenty of follow-up questions before starting to work. L3-4 for agent operation (once configured, agents run autonomously in the background, only requesting user involvement when encountering blockers or ambiguity.
User approval requirements for different decision types: The Agent will reach out to the user for clarification in the case of ambiguity or error
Execution monitoring, traces, and transparency: Agent shows reasonably detailed CoT, plans, and actions
Emergency stop and shut down mechanisms and user control: User can pause/stop the agent at any time
Usage monitoring and statistics and patterns: Monitor specific agent runs (link, archived), view activity logs, and see run history. Zapier Agents provides an Activity log so you can “see how agents interact with your data,” serving as an audit/monitoring control (link, archived)

Ecosystem interaction

Identify to humans?: None found
Identifies technically?: None found
Interoperability standards and integrations: Zapier’s MCP server lets you select/manage “tools” (Zapier actions within apps) that are exposed to the connected AI client (link, archived)
Web conduct: - Zapier's web scraping guide mentions that users should obey robots.txt (link, archived)

Safety, evaluation & impact

Technical guardrails and safety measures: workspace-level access controls, and app controls for which apps and actions Agents in that workspace can be equipped to use. Actions are bounded by the permissions granted to each connected app account. [company source]
Sandboxing and containment approaches: None found
What types of risks were evaluated?: None found
(Internal) safety evaluations and results: None found
Third-party testing, audits, and red-teaming: None found
Benchmark performance and demonstrated capabilities: None found
Bug bounty programmes and vulnerability disclosure: None, general bug bounty (link, archived) lists agents.zapier.com as out of scope
Any known incidents?: None found
← Back to 2025 Index