How to
How to replace a read-only connector with full read-write
A read-only connector can read your data into a conversation but cannot change anything. Replacing it with a custom connector adds the write tools, so Claude can create and update records, not just summarize them.
To replace a read-only connector with full read-write, swap it for a custom MCP connector that adds create and update tools alongside the existing read tools. It authenticates as your own account, writes to your real schema, and keeps deletes and outbound sends behind a confirmation you approve.
Key takeaways
- A read-only connector only exposes read tools, so Claude can summarize but cannot change records.
- A custom connector adds scoped create and update tools to the same data, including custom fields.
- Create and update flow inside a conversation, while delete and send stay behind a human confirmation.
- The connector authenticates as you and never grants more access than your own account has.
- You can keep your read-only setup for staff who only need to look, and add read-write where it matters.
Why is my connector read-only in the first place?
Your connector is read-only because it was built or scoped to expose read tools only: search, list, and fetch. Those tools pull records into a conversation but include no create, update, or delete actions. Adding write means giving Claude tools that map to your system's real write operations, which a custom connector provides.
Many directory and off-the-shelf connectors lean read-only on purpose. Reading data is lower risk and covers the common case of asking questions, so the write side is often left out or limited.
The result is a connector that answers questions well but leaves the actual data entry with your team. Replacing it does not mean throwing the read side away: it means adding the write tools that were missing, mapped to the same records you already read.
How do I add write access to an existing connector?
To add write access, replace the read-only connector with a custom one that maps your system's create and update operations to tools Claude can call. We mirror your real schema, including custom fields and pipelines, so writes land in the same structure your team and reports already depend on, not a simplified copy.
In practice this is a swap, not a patch:
- 1Remove the read-only connector from Claude.
- 2Add the custom connector in its place.
- 3Re-authenticate so it acts as your account with your permissions.
Because the new connector covers both reading and writing, the conversations you already run keep working. The difference is that Claude can now move a record, update a field, or log an activity in the same place it used to only describe one.
What actions become available once it is read-write?
Once a connector is read-write, Claude can create records, update fields, advance stages, and attach notes inside a conversation you start. The lookups you already had stay, and the new write tools handle the changes that used to mean clicking through screens. You choose which of those actions run automatically.
Create and update are the everyday actions: add a record, fill in a custom field, change a status, append a note. These can flow inside the conversation without leaving it.
Delete and outbound send (emailing a contact, posting a message, removing a record) are separate gated tools. They wait for a human confirmation before they run, so the high-stakes actions always have you in the loop.
Irreversible actions stay gated
Deletes and outbound sends never run automatically. They pause for a confirmation you approve, so nothing you cannot take back happens without you signing off first.
Will replacing the connector change who can access what?
Replacing the connector does not widen access. The custom connector authenticates as your own account and inherits your permissions, so it can only touch what you can already touch. Read and write tools are separate, and the connector never grants Claude more access than the account behind it has.
You decide which actions are automatic and which require a confirmation step, so the move from read-only to read-write does not mean handing over unsupervised control.
If different people need different levels, you can keep the read-only setup for staff who only need to look and add the read-write connector where the work actually happens.
Does this work for a custom or in-house tool with no listing?
Yes. A custom or in-house tool that appears in no connector directory is exactly the case a custom connector handles. We build the connector for your system and schema, with scoped read and write tools, including tools that run on-prem behind a firewall. If it has an API or a database, Claude can read and write to it.