Getting Started

Welcome to TestIntel! This guide covers your first steps after installation.

1. Configure your AI provider

TestIntel needs an AI provider to generate tests. It's Bring-Your-Own-Key: you

supply the credentials, all from the UI.

1. Go to Settings → System and pick your AI Provider:

• Anthropic Claude (recommended)
• OpenAI
• AWS Bedrock (Claude models — for teams already on AWS)

1. Go to Settings → API Keys & Tokens and add the matching credentials:

• Anthropic Claude: anthropic_api_key
• OpenAI: openai_api_key
• AWS Bedrock: aws_access_key_id and aws_secret_access_key

Optionally set a specific Model in Settings → System (leave blank for the

default). If no AI key is configured, a warning banner appears after login.

Using AWS Bedrock? It has a few extra AWS-side steps (model access, region,

IAM permissions) — see the AWS Bedrock setup guide below.

2. Create your first project

Click + New Project in the Chat tab:

1. Enter a project name (e.g. "my-app")

1. Optionally add your GitHub repo URL (enables code context and impact analysis)

1. Optionally add your Jira project key (enables Jira-aware chat)

The project scopes your inventory, staging, and results so multiple projects stay organized.

3. Start chatting

Ask TestIntel to generate tests. Try prompts like:

• "Generate test cases for user login"
• "Create a test strategy for the checkout flow"
• "What tests are affected by PR #42?"
• "Write Playwright automation for the signup page"
• "Analyze coverage gaps in my API"

The AI uses your project's code context, knowledge base, and integrations to produce relevant output.

4. Review and manage tests

• Add to Staging — Click the button on AI responses containing test cases
• Staging tab — Review, edit, and promote tests to inventory
• Inventory tab — Your permanent test repository with filters, export, and sync

5. Connect integrations (optional)

In Settings → API Keys & Tokens, add credentials for:

• Jira — Pull requirements, link tickets to tests
• Xray / Zephyr Scale — Sync test cases bidirectionally
• GitHub — Enable code context for smarter test generation

After adding credentials, run schema discovery from the Project tab to auto-detect your instance's configuration.

6. Upload knowledge base documents

Add project-specific context to improve AI output:

• Go to the Project tab → Knowledge Base section
• Upload documents: test standards, architecture docs, business rules, API specs
• The AI references these on every chat message for that project

Tips

• Branch & PR fields — Fill these in to give the AI code context for a specific change
• Include inventory context — Check this box when generating tests to help the AI avoid duplicates
• Keyboard shortcut — Press Enter to send a message, Shift+Enter for a new line
• New Chat — Click "New Chat" to start a fresh session (clears conversation history)

AWS Bedrock setup

Use this if you want to run Claude through AWS Bedrock (e.g. your team is

already on AWS and prefers billing/data to stay in your AWS account) instead of

going to Anthropic directly. TestIntel supports Claude models on Bedrock.

There are two sides: a few one-time steps in AWS, then entering the details

in TestIntel — all from the UI.

A. In the AWS console (one-time)

1. Model access. Serverless foundation models are now enabled

automatically on first invocation — there is no longer a manual "Model

access" page to toggle. Two things still apply for Claude:

• First-time Anthropic use: you may be prompted to submit brief use-case

details before Claude can be invoked in your account. Do this once via the

Bedrock Model catalog / playground.

• Org restrictions: account admins can still limit models via IAM

policies or Service Control Policies. If access is blocked there, requests

fail with an "access denied" error even with valid keys.

1. Get a credential for TestIntel. Two options:

• Bedrock API key (simplest): Amazon Bedrock console → API keys →

generate a key. It's scoped to Bedrock only, needs no IAM user, and is the

easiest path. Console keys expire (e.g. 30 days) — use a long-term key for a

persistent instance. Skip step 3 (IAM policy) if you use this.

• IAM access key: In IAM, create (or reuse) a user, and under

Security credentials → Access keys, create an access key. "Application

running outside AWS" or "Local code" is an appropriate use case. You'll get an

Access key ID and a Secret access key (shown once — copy it).

1. Grant least-privilege permissions. Attach a policy allowing only what

TestIntel needs:

• bedrock:InvokeModel and bedrock:InvokeModelWithResponseStream (generation + streaming)
• sts:GetCallerIdentity (lets TestIntel's key-verification show "valid")

Avoid attaching broad admin permissions. **Note on cross-region inference

profiles:** current Claude models are invoked through a cross-region

inference profile (an ID like us.anthropic.claude-...). Invoking one calls

the underlying model across several regions, so the policy must allow the

invoke actions on both the inference-profile ARN and the

foundation-model ARNs it routes to. If you get "access denied" only from

TestIntel but the console playground works, this policy scope is the usual

cause — for a quick test you can widen the resource to *, then tighten.

1. Note your region (e.g. us-east-1, eu-west-1) — it must match the

region prefix of the model ID you use in TestIntel (step B.2).

B. In TestIntel (all in the UI)

1. Settings → System → set AI Provider to AWS Bedrock.

1. Set Model to the Claude model's exact ID from its details page in the

Bedrock console. Current Claude models require the **inference-profile ID with

a regional prefix** — e.g. us.anthropic.claude-sonnet-4-6 (US), or the

eu./apac. equivalent — that matches your Bedrock Region. The bare

anthropic.claude-... form (no prefix) will fail with "the provided model

identifier is invalid" for these models. Open the model in the Model catalog

and copy the ID it shows (look for the Inference profile ID).

(Only Claude models are supported. If you enter a non-Claude model, TestIntel

warns you and — if results are poor — recommends switching to a Claude model.)

1. Set Bedrock Region to the region from step A.4.

1. Settings → API Keys & Tokens → add credentials. There are two options

— pick one:

• Bedrock API key (simpler): add bedrock_api_key — a single key from the

Bedrock console (Amazon Bedrock → API keys). No IAM user needed; it's scoped

to Bedrock automatically. Note: console-generated keys expire (e.g. 30

days), so for a long-running instance use a long-term key and plan to rotate.

• AWS access keys: add both aws_access_key_id and aws_secret_access_key

(the IAM access key from step A.2). Better for governance/audit (IAM policies,

SCPs).

If both are set, the Bedrock API key takes precedence. If it fails (e.g.

expired), TestIntel reports that clearly — remove the API key to fall back to

your access keys. (The region is *not* a secret — it lives in System settings.)

1. Save. TestIntel verifies the credential; a valid status confirms it works.

Then open the Chat tab and generate some test cases to confirm.

Troubleshooting

• "the provided model identifier is invalid" — the model needs the

inference-profile ID with a regional prefix. Use e.g. us.anthropic.claude-...

(matching your region), copied from the model's details page — not the bare

anthropic.claude-... id (step B.2).

• "AWS account does not have access to model …" — for first-time Anthropic

use, submit the use-case details in the Bedrock Model catalog/playground once

(step A.1); or your IAM policy doesn't cover the cross-region inference profile

(step A.3 — widen to * to test); or the region prefix doesn't match your

Bedrock Region. If the console playground also errors on a prompt, it's an

account/region access issue, not a TestIntel or key-scope problem.

• "AWS credentials are invalid" — the access key/secret are wrong or the key

was deactivated. Re-check both values in API Keys & Tokens.

• Key status shows something other than "valid" — if your IAM policy omits

sts:GetCallerIdentity, the credentials may still work for Bedrock but can't be

verified; add that permission for a clean status.

• "boto3 … not installed" — the AWS SDK isn't present in your deployment;

install with pip install boto3 (or the [all] extra). The standard Docker

image includes it.

Need help?

• Click Help in the header for documentation on specific features
• Check the FAQ for common questions
• Contact support at support@octobluetech.com