Integration · Developer tools

Connect Claude & ChatGPT to Letta

Work with your Letta agents from the AI assistant you already use. Ask it what an agent remembers, change its persona or system prompt, attach a tool, or send the agent a message and read the reply. It can also browse conversations, message history, runs, folders and the models available to your agents.

Connect Letta free

Free plan, no credit card. Takes about a minute.

Try asking

“What does my support agent have in its human memory block?”
The AI finds the agent and reads its human core memory block.
“Search the research agent's archival memory for anything about pricing.”
It searches the agent's archival memory by meaning and returns the closest passages.
“Attach the web search tool to my support agent.”
It finds the tool ID in your account and attaches it to the agent.

What your AI can do

Agents and memory

List and read agents, update their name, system prompt, model and tags, and read or replace core memory blocks such as persona and human.

Archival memory

List, search, add and delete passages in an agent's long-term archival memory.

Messages, tools and runs

Send messages to agents or conversations, read message history, attach and detach tools, and check runs, folders and models.

“Ask the onboarding agent to summarize what it knows about Acme.”
It sends the message to the agent and returns its reply. The agent may call its own tools and update its memory while answering.
“Switch the triage agent to a different model.”
It lists the available models, then updates the agent with the one you pick.
“Did the agent's last run finish, and why did it stop?”
It lists the agent's runs and reads the status and stop reason of the latest one.

25 tools for Letta

These tools are switched on when you connect. You can switch any of them off, or require your approval before it runs.

add_archival_memoryChanges dataSave a passage to an agent's long-term archival memory: pass `agent_id` and `text`, optionally `tags`.
attach_toolChanges dataAttach a tool to an agent by `agent_id` and `tool_id` (tool-...), so the agent can use it.
create_conversationChanges dataStart a new conversation with an agent by `agent_id`, optionally with a `summary`. Use send_conversation_message to talk in it.
delete_archival_memoryDeletesPermanently delete one passage from an agent's archival memory by `agent_id` and `memory_id`. This cannot be undone.
detach_toolChanges dataRemove a tool from an agent by `agent_id` and `tool_id`. The tool itself is kept.
get_agentGet one agent by `agent_id` (agent-...), with its model, system prompt, tools and memory blocks.
get_memory_blockGet one core memory block of an agent by `agent_id` and `block_label` (for example human or persona).
get_runGet one run by `run_id`, with its status and stop reason.
list_agent_messagesList the message history of an agent by `agent_id`, optionally filtered by `conversation_id`.
list_agent_toolsList the tools attached to an agent by `agent_id`.
list_agentsList agents, optionally filtered by `name`, `query_text`, `tags` or `project_id`. Use this to find an agent ID before other calls.
list_archival_memoriesList passages in an agent's long-term archival memory by `agent_id`, optionally filtered by text with `search`.
list_conversation_messagesList the messages in one conversation by `conversation_id`.
list_conversationsList conversations, optionally for one `agent_id` or matching text in their summaries with `summary_search`.
list_folder_filesList the files in a folder by `folder_id`; set `include_content` to true to include file text.
list_foldersList data folders (sources) of documents, optionally filtered by `name`.
list_memory_blocksList the core memory blocks (such as persona and human) of an agent by `agent_id`.
list_modelsList the language models available for agents, optionally filtered by provider.
list_runsList agent runs, filtered by `agent_id`, `conversation_id`, `statuses` or `active`. Use to check on background work.
list_toolsList the tools available in your Letta account, optionally filtered by `name` or `search`. Use to find a tool ID to attach to an agent.
search_archival_memorySearch an agent's archival memory by meaning: pass `agent_id` and a `query`, optionally `tags`, `top_k` or a date range.
send_agent_messageVisible to othersSend a message to an agent and get its reply: pass `agent_id` and the text in `input`. Set `streaming` to false. The agent may call its tools and update its memory while answering.
send_conversation_messageVisible to othersSend a message in a conversation and get the agent's reply: pass `conversation_id` and the text in `input`. Set `streaming` to false.
update_agentChanges dataUpdate an agent by `agent_id`: change its `name`, `description`, `system` prompt, `model`, `tags` or `tool_ids`. Only the fields you pass change.
update_memory_blockChanges dataChange an agent's core memory block by `agent_id` and `block_label`: pass the new `value`, and optionally `description` or `limit`. The new text replaces the old.

Set up in three steps

1
Pick the app
Create a free PipMCP account and choose Letta from the app list.
2
Paste your key
Paste a Letta API key. Log in at platform.letta.com, open API keys, create a key and copy it.
3
Add the link to your AI
You get a personal MCP link. Add it to Claude, ChatGPT or Cursor:
1. Click your name, then Settings › Connectors › Add custom connector.
2. Paste your link as the Remote MCP server URL.
3. Switch it on from the + menu in a chat.

Questions

What can the AI do in Letta?

The AI can list and read agents, memory blocks, archival memory, conversations, messages, runs, tools, folders and models. It can update agents and memory blocks, add and delete archival memory, attach and detach tools, start conversations and send messages to agents.

Does the AI see my Letta API key?

No. Your Letta API key and settings are encrypted at rest and never shown to the AI. After you save them, they are not shown again, not even to you. The AI only sees the results of the tools it calls.

Can I control what the AI is allowed to do?

Yes. You choose which tools are switched on, so you can start read-only. Sending a message makes the agent act and possibly change its memory, and deleting archival memory is permanent. These actions can require your approval before they run. Every tool call is logged.

Does it work with ChatGPT?

Yes. In ChatGPT go to Settings › Apps & Connectors › Advanced and turn on Developer mode, then add your PipMCP link. Developer mode needs a paid ChatGPT plan: Plus, Pro, Business or Enterprise. The same link also works in Claude (Settings › Connectors › Add custom connector), Cursor and other MCP clients.

What does it cost?

PipMCP has a free plan with no credit card required. Paid plans bill per completed task. You also need your own Letta account.

What is the difference between core memory and archival memory?

Core memory blocks, such as persona and human, are short texts the agent always has in view. The AI can read a block or replace its text. Archival memory is a long-term store of passages the agent searches when needed. The AI can list, search, add and delete those passages.