evoclaw

Self-Evolving Ai Agent Wrapper. Powered by @thinkymachines

[evoclaw @evoclaw_]

most "memory" in AI agents is a lie.

it's just notes in a text file.
inject it into the prompt. call it "memory."
same model. same weights. forever.

EvoClaw is different.

every conversation actually changes the model.
not the context. not the prompt.
the weights.

LoRA fine-tuning runs in the background.
the model that answers on day 7
is literally not the same model as day 1.

that's not memory.
that's evolution.

and now it evolves inside your community.

📱 Telegram — drop it in your group. it learns from every message.
💬 Discord — slash commands, per-channel memory, live embed stats.

the more people talk, the faster it evolves.

your community is the training data.

→ https://t.co/pfM7MD5kgc

[evoclaw @evoclaw_]

most AI agent interfaces lock you to one LLM.

EvoClaw doesn't.

chat with your self-evolving agent using:
• Groq (free)
• OpenAI
• Anthropic
• OpenRouter
• Together AI
• EvoClaw Server

one interface. any model. agent learns from every conversation regardless of provider.

your API keys?

stored in YOUR browser only.
never sent to EvoClaw servers.
never logged. never shared.

you talk directly to the LLM provider.
EvoClaw is just the evolution layer in between.

the best part?

every message → scored by PRM
every good response → permanent skill
every failure → auto-evolved

your agent gets smarter no matter which LLM you use.

try it free → https://t.co/gTHgANl4Dx

[evoclaw @evoclaw_]

What is EvoClaw?

A wrapper that sits between your AI model and your users.

Every conversation:
→ Intercepted
→ Scored by reward model
→ Skills injected in real time
→ LoRA training submitted to cloud
→ Weights hot-swapped. Zero downtime.

Your agent evolves while it runs.

[evoclaw @evoclaw_]

EvoClaw trains your model automatically — no GPU needed.

Every 32 conversations → LoRA job submitted to Tinker cloud → weights hot-swapped. Zero downtime.

Kimi-K2.5 (~200B MoE). No GPU. No downtime. No manual steps.

https://t.co/Z3NYfrZ718

[evoclaw @evoclaw_]

EvoClaw wraps any AI model and makes it learn from your real conversations — automatically.

3 commands to get started:

- pip install evoclaw
- evoclaw init
- evoclaw start

No GPU needed. Cloud LoRA via Tinker handles all the training

⭐https://t.co/WOO1lJYeit

[evoclaw @evoclaw_]

Most AI agents stay the same.

EvoClaw is different - it evolves with you.

Automatically. Every conversation.