The server itself exposes 39 tools, and some of them can write to your Coda account. Behind TrustGate an agent only sees the tools you allow, requests and responses are inspected, and every call is logged. Anything outside the allowlist is denied before it reaches Coda.

How is this different from connecting Coda MCP directly?

Connected directly, an agent sees every tool the server exposes and holds the credential itself. Through TrustGate you choose which tools each agent gets, and every call is logged against the agent that made it.

What if the Coda MCP API changes?

We track the Coda MCP schema, so the tool catalogue stays current. Tools you have already allowed keep working, and anything new stays denied until you enable it.

Can I give different agents different Coda tools?

Yes. Access is set per agent: one can get a handful of read tools, another the full set. You grant the whole MCP or single tools within it.

Where can I see what my agents did in Coda?

Every call is logged against the agent that made it, with the tool it used. Activity can be sent on to your own monitoring.

Do I need to run the Coda MCP server myself?

No. TrustGate connects to the Coda MCP server and sits in front of it. Your agent only talks to your TrustGate URL, never to Coda directly.

Connect Coda MCP to your AI agents, securely.

One secured endpoint for every Coda call your AI agents make, scoped per user, inspected, and logged.

Get started for free

Get a demo

Try Coda MCP now

What your agents can ask once Coda MCP is connected.

Select the tools your AI agents can call

39 tools from the Coda MCP schema.

Showing 39 tools

searchreadSearch for content across documents, pages, tables, and rows, returning ranked results with URLs.
url_convertreadConvert between coda:// resource URIs and Coda web URLs.
whoamireadGet the current authenticated user information (ID, name, email) and the workspaces they belong to.
tool_guidereadGet detailed guidance before working in specific areas.
name_matchreadMatch approximate page, table, view, or person names to candidates in the document.
document_createwriteCreate a new Coda document.
document_deletehigh riskPermanently delete a Coda document.
document_outlinereadOutline a document: document metadata (title, icon, createdAt, owner, sourceDoc) plus a paginated list of its pages, without body content.
document_managehigh riskChange a document's access control (add, update, or revoke a permission) or move the document to a different folder. Requires share access; moving also requires edit access.
page_createwriteCreate a new page, optionally nested under a parent page.
page_updatewriteUpdate an existing page's properties and position in the page hierarchy.
page_deletehigh riskDelete a page from a document, optionally including its child pages.
page_describereadDescribe page metadata (title, pageType, isHidden, parentPageUri, position, icon, coverPhoto, subtitle) and its child pages.
allow_locked_page_editswriteRecord that this connection may edit locked pages. Call only after the user agrees. A later write in this connection then bypasses the lock. The lock stays on the page.
table_createwriteCreate a new table on a page with specified columns and optional initial data.
table_deletehigh riskDelete an entire table from a document.
table_rows_readreadRead row data from a table with cell values, with filtering and pagination.
table_rows_managewriteAdd or update table rows.
table_rows_deletehigh riskDelete rows from a table, either specific rows or all rows.
table_columns_managewriteAdd or update table columns (name, type, formula).
table_columns_deletehigh riskDelete columns from a table.
table_view_managewriteAdd a new view to a table or configure an existing view, including charts, filters and sorts.
table_view_deletewriteDelete a table's view.
table_columns_readreadRead a table's column schema: column names, IDs, isWritable flag, format.type, formula bodies, and rowCount.
schema_explainreadMap a document's tables, columns, views, and the lookups between them. Returns each node's uri, name, and kind.
content_modifywriteModify canvas content: insert, replace, or delete elements.
content_duplicatewriteDuplicate content from one location to another, copying pages or documents to new or existing documents.
content_image_uploadwriteUpload an image to fill a previously inserted image placeholder.
content_readreadRead body content of a page or cell canvas.
document_suggest_changeswritePropose tracked suggestions on line content of a page or a canvas cell. A reviewer accepts or rejects each suggestion. This does not apply a live edit.
comment_addwriteAdd a comment or reply to a page element, table row, or existing thread.
comment_deletehigh riskPermanently delete a comment reply or an entire comment thread.
comment_resolvewriteMark a comment thread as resolved, or reopen a resolved thread.
notifications_listreadList the current user's notifications (mentions), newest first. Defaults to unread.
notifications_mark_readwriteMark notifications read by the mention timestamps returned by notifications_list. Mark only the notifications the user selected.
formula_executehigh riskEvaluate a Coda Formula Language (CFL) expression and return the result.
control_managewriteCreate or update a canvas control (formula or button).
control_deletewriteDelete a control from the document.
control_readreadRead a canvas control's full configuration by its formula or control URI — any control type, not just formulas and buttons.

What Coda MCP exposes

Coda MCP groups its 39 tools into 7 categories, each with a suggested starting policy.

All 39 tools as JSON, ready to paste.

Category	Covers	Access	Suggested policy
Search	Searching across documents, pages, tables and rows, converting between coda:// URIs and web URLs, returning the authenticated user and their workspaces, fuzzy-matching page, table, view or person names, and in-server guidance	Read	Read-only and the starting point for almost every task. Search results reach anything the connected account can open, so the account itself is the real boundary
Documents	Creating a new document, outlining a document's metadata and page structure, and permanently deleting a document	Read / Write	document_delete is permanent, so require approval for it. Coda's Admin API lets workspace admins see which documents were created or deleted through MCP
Pages	Creating pages, optionally nested under a parent, updating page properties and hierarchy, describing page metadata and child pages, and deleting a page with its children	Read / Write	page_delete can take child pages with it, so it belongs behind approval. Creating and updating pages are routine edits worth logging by document
Tables	Creating and deleting tables, reading, adding, updating and deleting rows, reading, managing and deleting columns, and adding, configuring and deleting views with charts, filters and sorts	Read / Write	This is where bulk changes happen. Row, column and table deletes remove data and should require approval, and Coda's own guidance is to require human approval for bulk table changes
Content	Reading the body of a page or cell canvas, inserting, replacing or deleting canvas elements, duplicating pages or documents into new or existing documents, and uploading images into placeholders	Read / Write	content_read pulls document text into the agent's context, which is where prompt injection from untrusted content enters. content_modify can delete elements as well as add them
Comments	Adding comments and replies to page elements, rows or threads, resolving and reopening threads, and permanently deleting a reply or a whole thread	Read / Write	Comments are visible to collaborators, so adding them speaks on the user's behalf. comment_delete is permanent and worth gating
Formulas & Controls	Evaluating a Coda Formula Language expression and returning the result, and creating, updating or deleting formula and button controls on a canvas	Read / Write	formula_execute evaluates arbitrary formula expressions against the document, so the expression text is what belongs in the log. Buttons created through control_manage can run actions when pressed

Connect Coda MCP to your AI agent

Four steps, under ten minutes.

01
Sign in to NeuralTrust
Open the NeuralTrust platform, sign in and go to Getting started, then Connect your tools.
02
Choose your agent
When asked which agent you use, pick Claude Code, Codex, Cursor, Windsurf, Copilot or your own code.
03
Connect Coda
Find Coda in the tool list and connect it.
04
Connect your agent
Follow the steps on screen to add the config to your agent, then send the test prompt.

Permissions down to the single tool

You do not allow Coda MCP or block it. You allow the exact tools each agent can call, one by one.

Coda MCP	39 TOOLS AVAILABLE
Code review agent	4 tools, read only
Issue triage bot	9 tools
Platform team assistant	all 39 tools

Tool by toolConnecting the server is not the unit of access. Each tool is granted on its own.
Default denyA tool you have not enabled is never exposed to the agent.
Approval on writesTools that change or delete things can require a human to confirm.

Is Coda MCP secure?

One endpoint instead of raw credentials. Every call is checked against policy and logged.

AI agentsConnect through your MCP client configuration.	NeuralTrustApply controls based on your configuration.	Coda MCP39 documented tools on the remote server at coda.io/apis/mcp over HTTP, in beta. OAuth 2 with PKCE currently grants only the mcp:all scope, so access is read and write; personal API tokens restricted to MCP can be read only, write only, or read and write. Editors are limited to read-type tools.
Coding assistants	Connection & credentialsper user	Tables11 tools
Internal copilots	Tool access policiesdeny by default	Search5 tools
Workflow automations	Request & response inspectioninline	Pages5 tools
Compatible MCP clients	Activity monitoringalerts and logs	All other tools18 tools