Project Configuration

A FireGuard project defines what your API requests actually run. Configure the project first, then use its project ID to create conversations and execute guardrails.

Project ID

Find the project ID in Project Settings → General. Pass it as project_id when creating conversations:

curl -X POST \
  "https://api.fireraven.ai/public/fireguard/v1/conversation?project_id=$FIREGUARD_PROJECT_ID" \
  -H "X-Api-Key: $FIREGUARD_API_KEY"

You can regenerate the project ID from project settings, but doing so invalidates the previous ID. Update deployed integrations before rotating it.

FireGuard Agent Hooks use this value as FIRERAVEN_PROJECT_ID in each installed hooks directory.

Policies Guardrail

The Policies Guardrail tab attaches organization policies to the project. For each policy, choose whether it applies to:

• Input guardrails, to evaluate user messages before your model runs.
• Output guardrails, to evaluate assistant responses before users see them.
• Both directions.

The project execution mode, normal or fast, controls the default runtime behavior for policy checks unless an API request overrides it with execution_mode.

Security Guardrail

The Security Guardrail tab configures jailbreak, prompt injection, and similar attack detection.

Key controls:

• Apply to input and Apply to output decide which directions run security checks.
• Sensitivity controls how strict the detector is.
• Security violation message is the fallback text your app can return when a security check blocks a request.

When images[] are sent to input_guardrails, security uses OCR and metadata text with conversation history. It does not receive raw image bytes.

Topics

Topics provide relevance monitoring and analytics. They do not block requests by themselves; blocking comes from policies and security guardrails.

On the Topics tab:

• Add organization topics to the project.
• Enable input analysis, output analysis, or both.
• Remove a topic from this project without deleting the organization topic.

Try It Out Sandbox

Use Try it out from project settings to test the project in a sandbox before deploying an integration. The sandbox uses the same guardrail configuration and request-unit rules as the public API.

Create a test conversation, send representative messages, and verify whether policy, security, and topic behavior matches your intended production flow.

Create Projects By API

You can create projects programmatically when you need repeatable environment setup or customer onboarding automation.