Fibery

Query a Fibery workspace, create and update entities, and reshape the workspace schema itself.

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

What the Fibery connector does

The Fibery MCP server connects your Fibery workspace to Claude, ChatGPT and any MCP-compatible agent. It exposes 55 tools that read and query your data, create and update entities and documents, and — unusually — rebuild the workspace schema itself: new spaces, new databases, new fields and new views. Fibery hosts it, sign-in is OAuth 2.0, and nothing installs locally.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Fibery tools (55)

  • add_collection_items
  • add_comment
  • add_file_from_url
  • append_document_content
  • create_avatars_fields
  • create_comments_fields
  • create_databases
  • create_entities
  • create_files_fields
  • create_formula_field
  • create_icon_fields
  • create_multi_select_fields
  • create_primitive_fields
  • create_relation_fields
  • create_single_select_fields
  • create_space
  • create_view
  • create_workflow_field
  • delete_avatars_fields
  • delete_comments_fields
  • delete_databases
  • delete_entities
  • delete_fields
  • delete_icon_fields
  • delete_space
  • delete_views
  • delete_workflow_field
  • download_file
  • fetch_by_url
  • fetch_view_data
  • get_connectors_list
  • get_documents_content
  • get_entity_links
  • get_files_meta
  • get_manual_import_link
  • get_me
  • get_tool_reference
  • query
  • query_views
  • remove_collection_items
  • rename_databases
  • rename_fields
  • schema
  • schema_detailed
  • search
  • search_guide
  • search_history
  • set_document_content
  • set_state
  • update_entities
  • update_formula_field
  • update_multi_select_fields
  • update_single_select_fields
  • update_view
  • update_workflow_field

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • The tool names here come from Anthropic's directory listing, not from a handshake. The server is OAuth-gated and we hold no Fibery credentials, so we did not read tools/list. Every read, write and structure label on this page is therefore derived from Fibery's documentation and from the names themselves — not from readOnlyHint or destructiveHint annotations, which we could not see. Treat any tool as capable of writing unless Fibery's guide says otherwise.
  • Fibery's documented tool set has moved past the directory snapshot. Its tools reference documents a block-level document editing family (read_document, insert_document_blocks, replace_block_text and siblings), permission-inspection tools (display_schema_capabilities, display_entity_capabilities_via_sharing), a reports group and a custom-apps group — none of which appear in the 55 names Anthropic listed on 2026-09-08. In the other direction, the directory lists get_documents_content, append_document_content, set_document_content and get_tool_reference, which the current tools reference does not. Both sources were read on 2026-09-08; the live server is the tiebreaker and we could not read it.
  • The connector acts as you. Fibery's FAQ states the server calls the API with your own user token rather than a service token, so it inherits exactly your permissions — and your workspace's IP restrictions. Fibery adds that there is currently no way to allowlist a specific MCP client's IP addresses separately, because they are not guaranteed to stay stable.
  • Rich text editing is partial. Table structure edits, multi-column layouts, embeds and file attachments inside rich text are unsupported, per Fibery's FAQ.
  • No Deep Research or Agent mode. Fibery deliberately omits the generic search and fetch tools those Claude modes require.
  • Query quality is the user's job. Fibery's FAQ advises naming the databases, fields and filters you want when a query is slow or wrong, and suggests packaging a working session into a reusable document so the assistant can repeat it.
  • Session errors are a known failure mode. Fibery documents a "no transport found for session id" error and prescribes re-adding the connector, updating the client, then restarting it — a WebSocket session can fail to initialise on first connect.
  • We did not verify rate limits or plan gates. Fibery's MCP documentation names no rate limit and no plan requirement, and we did not authenticate to observe one. The 401 response carried no x-ratelimit-* headers. Absence of a documented limit is not evidence there is none — Fibery's general API and workspace limits still apply.

Frequently asked questions

Can the Fibery MCP server change my workspace structure, not just my data?

Yes. Fibery's own guide lists schema modifications as a third capability class alongside reads and writes, covering the creation and deletion of spaces and databases, fields of any type, workflow stages and views. Twenty-eight of the 55 tool names in Anthropic's directory listing operate on structure rather than records, so an agent here can alter the container your data lives in.

Why am I getting a 401 Unauthorized error from the Fibery MCP server?

IP restriction on your workspace is the most likely cause. Fibery's FAQ explains that its MCP server calls the Fibery API using your own user token rather than a service-to-service token, so a workspace with IP restriction enabled blocks the request unless it arrives from an allowed address — even for admins. The address that matters is the one your MCP client sends from, not your laptop's.

Can I connect more than one Fibery workspace to Claude?

Yes, using a documented workaround. Fibery's FAQ says to append a query parameter such as a workspace name to the end of the connector URL before adding it, once per workspace. The parameter does nothing on Fibery's side; it only makes each URL distinct enough that Claude registers a separate connector instead of merging it with one you already added.

Why does Claude say the Fibery connector cannot be used in Deep Research or Agent mode?

Because Fibery does not implement the two tools those modes require. Its FAQ states the MCP server omits the generic search and fetch tools that Deep Research and Agent modes depend on, on the grounds that they do not fit typical Fibery use cases. Fibery says the server is instead built for data querying and reporting, entity creation and updates, and workspace automation.

What can the Fibery MCP server not do with rich text and documents?

Four things, per Fibery's FAQ: table structure edits such as moving rows and columns, creating two-, three- or four-column layouts, embeds, and file attachments inside rich text. Fibery describes MCP document editing as updating small parts of documents and rich text fields, and notes the server can also add comments to rich text for a human to review afterwards.

Does the Fibery MCP server work outside Claude?

Yes. Fibery's installation guide documents setup for ChatGPT, Le Chat, Claude Code, Codex, OpenCode, AugmentCode, n8n, Make.com, Cursor, Windsurf, VS Code with GitHub Copilot and Zed, and states the server should work in any MCP client. OpenCode is the exception that needs a bridge: Fibery documents a local mcp-remote wrapper because OpenCode does not yet support remote streamable HTTP.

Sources

Use it in an agent

Put Fibery to work.

Connect Fibery once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Call (650) 285-1019Our AI receptionist answers.