Turn 1:  ████ 700 tokens ($0.002)
Turn 5:  ████████████████ 4,300 tokens ($0.013)
Turn 20: ████████████████████████████████████████ 12,500 tokens ($0.038)
                                              ↑ You're paying for THIS every call

git clone https://github.com/woling-dev/promptthrift-mcp.git
cd promptthrift-mcp
pip install -r requirements.txt

# Install Ollama: https://ollama.com
ollama pull gemma4:4b

{
  "mcpServers": {
    "promptthrift": {
      "command": "python",
      "args": ["/path/to/promptthrift-mcp/server.py"]
    }
  }
}

{
  "mcpServers": {
    "promptthrift": {
      "command": "python",
      "args": ["/path/to/promptthrift-mcp/server.py"]
    }
  }
}

Q: Can I drink olive oil straight?
A: Yes! Our extra virgin is drinkable. We have 500ml and 1000ml.
Q: What's the difference between PET and glass bottles?
A: Glass is our premium line. 1000ml PET is for heavy cooking families.
Q: Which one do you recommend?
A: For drinking: Extra Virgin 500ml. For salads/cooking: 1000ml.
Q: I also do a lot of frying.
A: For high-heat frying, our Pure Olive Oil 500ml (230°C smoke point).

[Compressed history]
Customer asks about olive oil products. Key facts:
- Extra virgin (500ml glass) for drinking, single-origin available
- 1000ml PET for cooking/salads (lower grade, family-size)
- Pure olive oil 500ml for high-heat frying (230°C smoke point)
[End compressed history]

You: "Pin the fact that this customer is allergic to nuts"

→ promptthrift_pin_facts(action="add", facts=["Customer is allergic to nuts"])
→ This fact will appear in ALL future compressed summaries, guaranteed.

Before (every API call sends ALL of this):
┌──────────────────────────────────┐
│ System prompt      (500 tokens)  │
│ Turn 1: user+asst  (600 tokens)  │  ← Repeated every call
│ Turn 2: user+asst  (600 tokens)  │  ← Repeated every call
│ ...                              │
│ Turn 8: user+asst  (600 tokens)  │  ← Repeated every call
│ Turn 9: user+asst  (new)         │
│ Turn 10: user      (new)         │
└──────────────────────────────────┘
Total: ~6,500 tokens per call

After PromptThrift compression:
┌──────────────────────────────────┐
│ System prompt      (500 tokens)  │
│ [Pinned facts]      (50 tokens)  │  ← Always preserved
│ [Compressed summary](200 tokens) │  ← Turns 1-8 in 200 tokens!
│ Turn 9: user+asst  (kept)        │
│ Turn 10: user      (kept)        │
└──────────────────────────────────┘
Total: ~1,750 tokens per call (73% saved!)