How do I sign up for Future AGI Starter and create my first project?

Most teams hit the same wall at the exact same moment: you move beyond a demo, LLM behavior starts drifting, and you realize you need a system to evaluate, debug, and improve agents—not just call models. That’s where Future AGI Starter comes in. It gives you a free, structured way to build your first evaluation loop: datasets → experiments → evals → improvements → monitoring.

This guide walks you step by step through how to sign up for Future AGI Starter and create your first project so you can start shipping more reliable AI, not just better demos.

Quick Answer: Create a Future AGI Starter account from the sign-up page, complete the basic onboarding, then spin up your first project from the dashboard by defining a dataset, setting up an experiment, and connecting your LLM or agent stack. You can start for $0 and begin evaluating an agentic workflow in minutes.

The Quick Overview

What It Is: Future AGI Starter is the entry tier of the Future AGI platform—a free, developer-first environment to build, evaluate, and improve LLM agents using synthetic datasets, experiments, and deterministic evaluations.
Who It Is For: Founders, ML engineers, and AI teams who want to move from ad‑hoc prompt tweaking to a structured evaluation loop—especially for RAG apps, multi-step agents, and early-stage AI products.
Core Problem Solved: LLMs are probabilistic. Starter helps you tame that variability with traceable experiments, measurable metrics, and repeatable evaluation, all without upfront cost.

How It Works

At a high level, Future AGI Starter helps you do three things in sequence:

Create an account and workspace so your team has a shared place for experiments, datasets, and traces.
Create your first project to evaluate a specific agent, workflow, or model configuration.
Run an experiment and review results so you can see exactly how your agent behaves and where it fails.

Under the hood, the same lifecycle Future AGI uses for larger teams is already there in Starter:

Datasets: Create or import synthetic and real-world test cases, including edge cases.
Experiment: Configure which models, prompts, or workflows you want to compare.
Evaluate & Improve: Run deterministic evaluations, pinpoint root cause with feedback, and iteratively refine prompts and workflows.

You can connect to your existing stack (OpenAI, Anthropic, Bedrock, Gemini; LangChain, DSPy, CrewAI, LiteLLM, etc.) using lightweight instrumentation, then treat each change as a measurable experiment—not a guess.

Step 1: Sign Up for Future AGI Starter

Future AGI Starter is designed so you can get going in minutes.

1. Go to the Future AGI sign-up page

Visit: https://futureagi.com
Use the Sign up or Create account CTA in the header.
If you’re a startup, you’ll also see the Future AGI for Startups banner (“Get 6 months of Pro access free plus $5,000 in credits. Apply Now!”). You can:
- Start with Starter immediately, and
- Optionally apply to the startup program for extended Pro access and credits.

If you have questions during signup, you can contact the team at startups@futureagi.com.

2. Create your account

You’ll be prompted to provide:

Name
Work email
Password (or SSO, if available)
Role / Company info (optional depending on the current flow)

After submitting, you’ll typically:

Verify your email (check your inbox/spam).
Log in to your new Future AGI account.

Once you’re in, you’re on the Starter plan by default (the “$0 forever (seriously)” tier in their pricing language), which is ideal for your first project and early experiments.

3. Complete basic onboarding

On first login, you’ll usually see a short onboarding or welcome flow. Expect steps like:

Set your primary use case
(e.g., “Evaluate RAG chatbot,” “Improve summarization quality,” “Test tool-using agent.”)
Choose your preferred providers
(e.g., OpenAI, Anthropic, Bedrock, Gemini).
Optional: Join the Future AGI Community
There’s a public community and Slack where teams share evaluation practices and get help.

Once onboarding is done, you land in the dashboard and are ready to create your first project.

Step 2: Create Your First Project

Think of a project as a dedicated environment to evaluate and improve one AI application (a RAG bot, a summarizer, a voice agent, etc.).

1. Start a new project

From the dashboard:

Click “Create Project” or a similar CTA (e.g., “New Project”).
Give your project:
- Name – e.g., Support RAG Bot, Sales Email Writer, Voice Agent QA.
- Description – a brief note on what you’re evaluating and why.

Example:

Name: Customer Support RAG v1
Description: Evaluate retrieval quality and hallucination rate across different RAG configurations for support tickets.

2. Select your core workflow type

Future AGI Starter is flexible, but you’ll get the most value by framing the project around a clear workflow type:

RAG Chatbot – Evaluate answer correctness, grounding, and hallucinations.
Summarization – Evaluate coverage, faithfulness, and style.
Tool-using / Multi-step Agent – Evaluate task success, tool accuracy, and latency.
Content Generation – Evaluate quality, safety, consistency with brand voice.

Choose the option closest to your use case. This influences recommended metrics and dataset patterns.

Step 3: Set Up Your First Dataset

LLMs are probabilistic. To evaluate them, you need a stable test set that you can reuse as you iterate.

In Future AGI Starter, you’ll work in the Datasets stage first.

1. Create a dataset

Within your new project:

Navigate to Datasets.
Click “New Dataset”.
Provide:
- Dataset name – e.g., Support Ticket Eval Set v1.
- Dataset description – what scenario it covers and why.

2. Populate the dataset

You have two main paths:

Import existing examples
- Upload a CSV/JSON (depending on current support) of:
  - input (e.g., user query, document)
  - Optional context (retrieved docs, metadata)
  - Optional expected_output or annotation fields
- Useful if you already have logs or historical tickets.
Generate synthetic examples
- Use synthetic data generation to:
  - Cover common paths (frequent user questions).
  - Capture edge cases (ambiguous queries, long contexts, adversarial prompts).
- Synthetic datasets help you stress-test behavior beyond what’s in your logs.

The goal: a reusable set of 20–200+ examples where “good behavior” and “bad behavior” are easy to spot.

Step 4: Configure Your First Experiment

With a dataset in place, move to the Experiment stage.

1. Choose what you’re testing

An experiment compares one or more configurations such as:

Different models (e.g., gpt-4.1 vs claude-3-opus).
Different prompts or system messages.
Different RAG pipelines (e.g., retrieval settings, chunk size, re-ranking).
Different agent workflows (tool order, decision policies).

For your first project, start simple:

Baseline: your current production configuration or best guess.
Candidate: a new prompt or new model you want to test.

2. Connect your model/provider

Future AGI integrates with:

OpenAI
Anthropic / Claude
Amazon Bedrock
Google Gemini
Hugging Face
Together AI

And it plays well with frameworks like LangChain, Haystack, DSPy, CrewAI, LiteLLM.

In Starter, you’ll typically:

Add a provider and API key in settings (if required by the current UI).
Select that provider and model in the experiment configuration.

3. Define experiment variants

Within the experiment:

Create Variant A (Baseline) – current prompt/model/pipeline.
Create Variant B (Candidate) – modified prompt/model/pipeline.

Example for a RAG bot:

Variant A: gpt-4.1 with existing system prompt and current retrieval.
Variant B: gpt-4.1-mini or new RAG chain with a revised system prompt emphasizing grounded answers.

Save the experiment configuration.

Step 5: Evaluate and Inspect Results

Now move to the Evaluate stage.

1. Select evaluation metrics

Future AGI emphasizes research-based, deterministic evaluations. For your first project, focus on:

Task quality metrics
- Accuracy / correctness
- Faithfulness to provided context (for RAG)
- Coverage (for summarization)
Safety metrics
- Toxicity / sexism
- Privacy / sensitive data leakage
- Prompt injection handling
User-experience metrics
- Coherence, style alignment, helpfulness

Starter will typically offer default metric bundles depending on your use case; you can often add custom ones as you mature.

2. Run the experiment

From the experiment view:

Click Run or Start Evaluation.
The system:
- Runs both variants across your dataset.
- Captures outputs, metrics, and traces.
- Stores everything for deterministic comparison.

You’ll see:

Per-example scores per variant.
Aggregate metrics showing which variant “wins.”
Traces to inspect model behavior step-by-step (especially for agents).

3. Pinpoint root cause

Use the evaluation results to answer:

Where does the candidate outperform the baseline?
Where does it regress?
Are failures tied to specific:
- input types (long docs, noisy queries),
- tools,
- or prompt sections?

Future AGI’s evaluation feedback helps you “pin-point root cause” rather than guess. That’s the foundation for the next step.

Step 6: Improve and Iterate

Evaluation without improvement is just a report. Starter is built to close the loop.

1. Refine prompts and workflows

Based on feedback:

Adjust your system prompt to fix specific failure patterns.
Tweak retrieval settings (RAG) based on grounding scores.
Reorder tools or change decision logic (agents) based on task success metrics.

2. Re-run experiments

Clone or update your experiment:

Create Variant C based on your improvements.
Re-run on the same dataset.
Confirm whether your changes actually improve metrics.

This is how you move from “I think this prompt is better” to “this variant improves accuracy by 25% on our evaluation set.”

Step 7: Monitor & Protect (When You Go to Production)

Once you’re confident in your best-performing configuration, the next step is to deploy it—and keep it under control.

In the full Future AGI lifecycle, that means the Monitor & Protect stage:

Monitor
- Trace production requests.
- Watch real-time metrics for drift and performance drops.
- Use logs to replay failures and add them back into your datasets.
Protect
- Apply guardrails based on categories like:
  - Toxicity and sexism.
  - Privacy and sensitive data.
  - Prompt injection and jailbreak attempts.
- Use minimal-latency filters (e.g., Protect) to block unsafe inputs/outputs in real time.

On Starter, you get a taste of this workflow; as you scale, Monitor & Protect becomes the backbone for continuous improvement and safety.

Features & Benefits Breakdown

Core Feature	What It Does	Primary Benefit
Datasets & Synthetic Data	Create/import reusable test sets, including edge cases.	Make LLM behavior measurable and comparable across iterations.
Experiments & Variants	Compare prompts, models, and workflows side by side.	Find the best-performing agent configuration with evidence.
Deterministic Evaluations	Apply research-based metrics and feedback loops.	Pinpoint root cause of failures and improve with confidence.

Ideal Use Cases

Best for early-stage RAG and agent projects: Because you can quickly validate whether a new prompt, retriever, or model actually improves answer quality before you ship.
Best for startups and lean AI teams: Because Future AGI Starter is $0 to begin with and integrates directly into your stack, so you get eval infrastructure without building it yourself.

Limitations & Considerations

Starter resource limits: As a free tier, Starter may cap dataset size, experiment count, or request volume. If you’re running heavy multimodal or large-scale evals, consider upgrading.
Production monitoring depth: Starter is ideal for building your first evaluation loop; advanced Monitor & Protect workflows and high-throughput guardrails are typically better served by Pro/Enterprise plans.

Pricing & Plans

Future AGI Starter is designed as a $0 forever entry point so you can build a serious evaluation loop without paying upfront.

Typical structure:

Starter (Free): Best for individuals and small teams needing to validate their first agent workflows, build datasets, and run core experiments.
Pro / Enterprise: Best for teams needing large-scale evaluations, multimodal workloads (text, image, audio, video), advanced Monitor & Protect, and tight integration with existing CI/CD and observability tooling.

If you’re a startup, you can apply to Future AGI for Startups to get:

6 months of Pro access free
$5,000 in platform credits

This is ideal if you expect to scale quickly past the Starter limits.

Frequently Asked Questions

Do I need a credit card to sign up for Future AGI Starter?

Short Answer: No, you can start on Future AGI Starter without a credit card.

Details: The Starter tier is intentionally frictionless. You create an account via the sign-up page, verify your email, and you’re ready to build datasets and run your first experiments. Payment details are only required if/when you choose to upgrade to a paid plan.

Can I connect my existing OpenAI/Anthropic setup to my first project?

Short Answer: Yes. Future AGI is designed to plug into providers like OpenAI and Anthropic and frameworks like LangChain, DSPy, CrewAI, and LiteLLM.

Details: In your project settings or experiment configuration, you can add your provider credentials (e.g., OpenAI API key) and select the models you want to test. Because the platform focuses on traces and experiments rather than replacing your stack, you can keep your existing orchestration (LangChain, DSPy, etc.) and use Future AGI as the evaluation and optimization layer.

Summary

Future AGI Starter gives you a structured, $0 way to move from LLM demos to reliable AI products. You:

Sign up for Starter and create your account.
Create your first project around a specific workflow (RAG, summarization, agent).
Build a dataset, run an experiment, and use deterministic evaluations to find a winning configuration.
Iterate by refining prompts and workflows and, when ready, extend into Monitor & Protect for production.

If you can’t measure and replay failures, you don’t have a product—you have a demo. Starter is how you build that measurement layer from day one.

Next Step

Get Started with Future AGI today, set up your Starter account, and create your first project to evaluate your agents with confidence.

Answers you can trust, from Codeables