Mimicry Telegram · Mimicry Agent
Your agent.
Inside your applications.
Use Mimicry in Telegram, or connect the agent you already use. Read chat history, retrieve context and reuse local embeddings without building those integrations from scratch.
Two ways into Mimicry
A real working setup
The setup I actually use.
Telegram, Hermes and my bots.
Connected through Mimicry.
Chat memory · selected account
Activity
What deadline did we finally agree on?
Return the decision
September 12
The later confirmation replaces the earlier proposal.
Mira · Launch team“Confirmed: September 12. QA needs the extra time.”
find_contact → search_chat → read_around
Real setup. Activity illustrated with sample messages.
How Mimicry works
From a conversation to an action.
Read the app, not a screenshot.
Telegram gives Mimicry chat identities, messages, senders and dates. The agent can request the right conversation directly instead of asking you to copy it into another assistant.
find_contact → read_recentBring back the part that matters.
Conversations are indexed as searchable passages. A question retrieves candidates from the chosen chat; nearby history lets the agent check a later confirmation against an earlier proposal.
search_chat → read_aroundCarry the result back into the app.
The native client delivers the prepared message. Hermes can keep orchestrating through MCP. Bots can reuse the embedding service while keeping their own messages and search indexes.
native actions · MCP · embeddings APIPlans · USD / month
The client is free.
Add the agent. Choose your compute.
Mimicry Telegram
Free
A Telegram client for your everyday conversations. No Agent subscription needed for ordinary client use.
Get the clientMimicry Agent
$3 / month
One personal Agent for supported Mimicry apps. Memory, MCP and local embeddings. Bring a supported provider or account connection; remote usage is separate.
Mimicry Compute
$20 / month
Let Mimicry handle the model usage. Compute includes Agent — not a second subscription on top.
Subscriptions are coming soon. Compute allowances and model availability will be published before payments open.
Already have an agent?
Keep it. Give it the context.
- Open External MCP in Mimicry.Use the command and arguments supplied by the client’s settings.
- Add that connection to your agent.Hermes or another MCP client stays in charge of the task. Mimicry supplies the native chat tools.
- Start with one conversation.Ask for recent messages or a decision in a chosen chat. Follow the source before acting on it.
Local memory is not an offline model.
Memory and embeddings run in the local Agent. A remote model still receives the context sent with its requests. Apps using the embeddings API keep their own search indexes.