# memory.write

Record a belief: subject, predicate, value, in a space. Optionally supersede an earlier fact. With evidence naming a fact the gateway fetched for this call, the value must equal it (or the field at path) and the write is verified; otherwise it is refused. Returns the new fact with its id, transaction time, actor, provenance, and source receipt.

Agent View of the PolicyLayer registry record for `memory.write`. HTML page: https://policylayer.com/tools/agent-custody/memory.write

## Facts

- Tool: `memory.write`
- Server: Agent Custody (`ch4r10t33r/agent-custody`) — https://policylayer.com/tools/agent-custody.md
- Homepage: https://github.com/ch4r10t33r/agent-custody
- Risk category: Write (Medium risk)
- Registry record: grade F, identity unverified
- Server rate-limited: no
- Parameters: 0
- Recommended policy verdict: Rate-limited

## Example call (MCP tools/call, JSON-RPC 2.0)

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "memory.write",
    "arguments": {}
  }
}
```

## Why memory.write is rated Medium

An AI agent can call memory.write faster than any human can review: one bad instruction and it creates or modifies resources in Agent Custody by the hundred, each call as confident as the last.

## Use case

AI agents use memory.write to create or update resources in Agent Custody, usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Agent Custody environment.

## Recommended policy (PolicyLayer)

Verdict: **Rate-limited**. Enforced by the PolicyLayer MCP gateway (https://policylayer.com/mcp-gateway) before a call reaches Agent Custody:

```json
{
  "version": "1",
  "default": "deny",
  "tools": {
    "memory.write": {
      "limits": [
        {
          "counter": "memory.write_rate",
          "window": "minute",
          "max": 30,
          "scope": "grant"
        }
      ]
    }
  }
}
```

## Other tools on Agent Custody (13)

- `memory.forget` — Destructive — https://policylayer.com/tools/agent-custody/memory.forget.md
- `memory.sweep` — Destructive — https://policylayer.com/tools/agent-custody/memory.sweep.md
- `customer.lookup` — Other — https://policylayer.com/tools/agent-custody/customer.lookup.md
- `echo.say` — Other — https://policylayer.com/tools/agent-custody/echo.say.md
- `memory.confirm` — Other — https://policylayer.com/tools/agent-custody/memory.confirm.md
- `memory.get` — Other — https://policylayer.com/tools/agent-custody/memory.get.md
- `memory.history` — Other — https://policylayer.com/tools/agent-custody/memory.history.md
- `memory.hold` — Other — https://policylayer.com/tools/agent-custody/memory.hold.md
- `memory.read` — Other — https://policylayer.com/tools/agent-custody/memory.read.md
- `memory.release` — Other — https://policylayer.com/tools/agent-custody/memory.release.md
- `memory.retract` — Other — https://policylayer.com/tools/agent-custody/memory.retract.md
- `stripe.payout` — Other — https://policylayer.com/tools/agent-custody/stripe.payout.md
- `stripe.refund` — Other — https://policylayer.com/tools/agent-custody/stripe.refund.md

## For agents

This record is a snapshot. Live verdicts and the full registry:

- Check every server in your MCP config at once: `npx -y policylayer stack`
- Vet a server before you add it: install the mcp-precheck skill — `npx skills add https://policylayer.com` (skill text: https://policylayer.com/skill.md)
- Query the registry over MCP: endpoint `https://api.policylayer.com/mcp` — tools `check_mcp_server`, `check_mcp_stack`, `check_tool`, `search_registry`, `get_change_events`

---

Source: the PolicyLayer MCP registry — one continuously verified record per MCP server. Full record: https://policylayer.com/registry?q=agent-custody · API: https://policylayer.com/registry/api · Policy library: https://policylayer.com/policies/agent-custody