Getting started
Use Campfire with Claude / GPT / Gemini / local models.
Campfire is the open-source AI customer service agent you can bring your own model to. Point Settings → AI Providers at your key, pick a model, and run. No per-resolution markup — you pay your provider directly; Campfire stays a flat workspace price.
Why BYO AI
Many support products charge a per-resolution AI tax on top of seats. Campfire does not. When you attach your own API key:
- Token spend bills at your provider rates (OpenAI, Anthropic via OpenRouter, Google, Groq, Together, or local Ollama).
- Campfire adds no per-resolution markup on those tokens.
- Workspace price stays flat (see pricing on the site) — no per-seat fees for AI access.
Keys are encrypted at rest and are write-only in the UI: after save you only see a masked form such as sk-••••1234, never the full secret again.
Configure a provider (settings UI)
- Open the dashboard and go to
Settings → AI Providers(/dashboard/settings/ai-providers). - Click Add Provider.
- Choose a provider preset (OpenAI, OpenRouter for Claude/GPT/Gemini, Groq, Together, Ollama, or custom OpenAI-compatible).
- Paste your API key (leave blank when editing if you want to keep the existing key).
- Pick or type a model id, optionally mark the provider as default, then save.
- Use Test Connection to hit the provider's
/modelsendpoint before going live.
Month-to-date token usage and resolution counts live under Settings → AI Usage.
Provider recipes
OpenAI (GPT)
Preset: OpenAI. Base URL https://api.openai.com/v1. Use a project API key from the OpenAI dashboard. Example models: gpt-4o, gpt-4o-mini.
Claude (Anthropic via OpenRouter)
Preset: OpenRouter (Claude / GPT / Gemini). Base URL https://openrouter.ai/api/v1. Create an OpenRouter key, then select a Claude model such as anthropic/claude-sonnet-4. Campfire talks OpenAI-compatible chat completions; OpenRouter routes to Anthropic.
Gemini (Google via OpenRouter)
Same OpenRouter preset. Choose a Gemini model id, e.g. google/gemini-2.0-flash-001. Alternatively use any gateway that exposes Gemini behind an OpenAI-compatible base URL.
Groq / Together
Pick the Groq or Together preset to fill the base URL, paste your key, and select a hosted open-weight model. Useful for low-latency or cost-sensitive tiers.
Local models (Ollama)
Preset: Ollama (local). Base URL defaults to http://localhost:11434/v1. API key is optional for local Ollama. Pull a model first, then set the model field to its name (e.g. llama3.3).
Example: OpenRouter + Claude
Values you enter in the dialog (keys stay in your browser only until save; the server encrypts them and never returns plaintext):
Provider: OpenRouter (Claude / GPT / Gemini)
Base URL: https://openrouter.ai/api/v1
API Key: sk-or-v1-... # write-only after save
Model: anthropic/claude-sonnet-4
Default: onHow runtime uses your config
When an AI reply is generated, Campfire loads enabled rows from your org's ai_providers table, decrypts the key server-side, registers an OpenAI-compatible adapter, and prefers the provider marked Default. If none are configured, env-based hosted providers remain the fallback.
Next: Quickstart to embed the messenger, or Widget for the full embed surface.