Blog

From the Prompt One team

Articles on enterprise agent architecture and compiled workflow agents.

aclif, the Agent CLI Framework, Is Open Source

Prompt One has released aclif, the framework under its Service CLI, as an MIT-licensed open source project. aclif builds command-line tools for AI agents: one grammar across every SaaS provider, canonical names, errors an agent can act on, introspection that needs no credentials, declared safety metadata, and an embeddable runtime for gateways.

Read the post →

Your Agent Should Never Hold a Credential

Part 2 of 2 on how a compiled workflow agent is built and run. When a deployed agent executes a command the Composer fixed at design time, the Service Gateway authenticates the calling agent, attaches the acting user's identity, checks scope, blast radius, entity and field rules, and query constraints, resolves the credential from the vault, and runs the command through the embedded Service CLI runtime. The agent never holds a credential and cannot widen its own scope.

Read the post →

Your Agent Should Never Choose Its Own Tools

Part 1 of 2 on how a compiled workflow agent is built and run. How the Composer authors an agent's commands at design time through a Service CLI that describes itself, why that loop belongs at design time, under review, and what it costs compared with an agent that loads MCP tool schemas on every run.

Read the post →

Why Deterministic Enterprise Automation Just Makes Sense

Sloppy analysis gets people fired, so why do we tolerate it from AI agents? Why I joined Prompt One, and what a platform built for business users rather than developers actually gives me.

Read the post →

Going Headless with Agentforce

How we built headless access to Agentforce agents for Prompt One's service gateway, first on Connected Apps and then through the migration to External Client Apps, plus the configuration details every third-party agent platform needs to get right.

Read the post →

What Happens When You Compile an Enterprise Workflow

Inside the compiled workflow agent architecture: the LLM as compiler, the agent workspace, security by construction, and a measured 11x cost advantage.

Read the post →

The Decision That Sinks AI Agent Projects Is Made Before the Project Starts

Why agent projects fail on an architecture choice most teams never consciously make, and the eight questions that make the choice deliberate.

Read the post →