The Infrastructure Layer for the OpenAI Agents SDK

Runner
real loop
50
MCP tools
5
SDKs, one layer
1 cmd
docker compose up

agentFast is the production layer for the OpenAI Agents SDK. The SDK gives you the Runner loop, function tools, and hand-offs. agentFast adds memory, observability, guardrails, evals, and durable human-in-the-loop around it — nothing about the production layer is a second-class citizen here.

What agentFast adds to the OpenAI Agents SDK

Real FunctionTools, real Runner

The tool registry is bridged as real FunctionTool instances and runs the real Runner loop — not a reimplementation.

Durable, crash-proof runs

Checkpoint-per-iteration to Postgres, so a killed process resumes from the exact step, no replay.

Observability upgrade

Beyond basic tracing: a full step tree with cost, latency, prompt-cache accounting, OTEL and LangSmith export.

Memory

Short-term buffer, long-term vector recall, episodic log and scratchpad, with compaction for long runs.

Guardrails

PII redaction before the model sees it, injection defense and token-bucket rate limits.

Durable human-in-the-loop

Gate risky tools; approvals execute exactly once via a database compare-and-set.

From clone to a running OpenAI Agents SDK agent
  1. 1
    Install with the OpenAI Agents extra
  2. 2
    Scaffold an agent
  3. 3
    Bring up the full stack
  4. 4
    Gate on quality
bash
pip install -e ".[openai-agents,dev]"
agentfast new support --sdk openai_agents_sdk
cd support-agent && docker compose up
agentfast eval
The SDK vs. agentFast on the SDK
OpenAI Agents SDKagentFast on the SDK
Runner loop, function tools, hand-offs✓ its core strengthUses the SDK directly
Memory (vector + episodic)— bring your own✓ Included
Observability + cost tracingBasic tracingFull step tree, OTEL, LangSmith
Guardrails (PII, safety, limits)— bring your own✓ Included
Durable human-in-the-loop— bring your ownSurvives a crash, exactly-once
Evals in CI— bring your own✓ Included
Working reference agentExamplesSupportAgent, point it at your KB
Pick the SDK, keep the layer
agentFast's seam holds across graph-native, hand-rolled, vendor-CLI, vendor-Runner, and role-crew models — switch orchestrators without rebuilding production plumbing. Explore evals, memory, and human-in-the-loop.

Read why the production layer is the real work, or see the comparisons.

Stop rebuilding the plumbing.

agentFast is the production layer — memory, 50 tools, observability, evals, guardrails, human-in-the-loop — for LangGraph, CrewAI, the Claude & OpenAI Agent SDKs & Vanilla. Own it for life.

Get agentFast — from $89
Keep reading
The Production Layer Every AI Agent Needs (and Why LangGraph Doesn't Give It to You)
9 min read
LangGraph vs CrewAI vs a Vanilla Loop: Which Should You Actually Use in 2026
11 min read
How to Add Human-in-the-Loop Approval to Any AI Agent (With Durable Execution)
10 min read