Agent Memory Block

AI agents forget context between sessions. Every new conversation starts blank. Your agent doesn't remember it uses CueAPI, what cues it has, or how to interact with the API.

The memory block solves this. It's a preformatted text block that your agent can save to persistent memory, so it always knows how to use CueAPI.

What's in the memory block

The memory block contains:

• Your masked API key (first 12 characters + ***)
• Your plan (Free, Pro, or Scale)
• Number of active cues
• List of active cues with human-readable schedules
• Dashboard and worker install links
• Instructions for common operations

How to enable it

The memory block is off by default. The user must explicitly consent:

curl -X PATCH https://api.cueapi.ai/v1/auth/me \
  -H "Authorization: Bearer cue_sk_your_key" \
  -H "Content-Type: application/json" \
  -d '{"memory_block_enabled": true}'

How to fetch it

Once enabled, fetch the memory block:

curl https://api.cueapi.ai/v1/memory-block \
  -H "Authorization: Bearer cue_sk_your_key"

If the user hasn't enabled it, you'll get a 403:

{
  "error": "memory_block_disabled",
  "message": "Agent memory block is disabled. Enable it via PATCH /v1/auth/me with memory_block_enabled: true."
}

How to save it

Claude (Projects)

Add the memory block text to your Claude Project's custom instructions.

OpenAI (Custom GPTs)

Paste the memory block into your GPT's system instructions.

ChatGPT (Memory)

Ask ChatGPT to "remember this" and paste the memory block.

Custom agents

Store the memory block in your agent's system prompt, persistent memory store, or configuration file. Refresh it periodically by calling GET /v1/memory-block.

Automatic updates

When you enable the memory block, CueAPI automatically creates a weekly cue called cueapi-memory-update-check on your account. Every Monday at 9am UTC it fires and instructs your agent to check for new CueAPI features. If the version is newer than what your agent has stored, it updates its memory automatically.

You say yes once. Your agent stays current forever.

For instant push notifications instead of weekly checks, subscribe:

curl -X POST https://api.cueapi.ai/v1/memory-block/subscribe \
  -H "Authorization: Bearer $CUEAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"callback_url": "https://your-agent.com/webhook"}'

Check usage before creating cues

The memory block includes a live usage summary. Agents should also call GET /v1/usage before creating new cues to avoid hitting limits:

usage = requests.get("https://api.cueapi.ai/v1/usage", headers=headers).json()
 
# Check if there's room for a new cue
if usage["cues"]["remaining"] <= 0:
    print("No cue slots available")
 
# Check if executions will exceed the limit
if usage["executions"]["will_exceed_limit"]:
    print(f"Projected {usage['executions']['projected_month_end']} executions by month end")

The usage response includes projected_month_end and will_exceed_limit so agents can make informed scheduling decisions without manual calculation.

Privacy

• Off by default: user must explicitly enable via PATCH /v1/auth/me
• Masked API key: only the first 12 characters are shown, never the full key
• Cue names only: no payloads, no callback URLs, no execution data
• User controls it: can disable anytime with memory_block_enabled: false
• No tracking: CueAPI does not track whether or how the memory block is used