OpenClaw Feishu Base#

Use this plugin when working with Feishu Base / Bitable in OpenClaw.

What it provides#

A unified tool:

• feishu_base

Recommended actions include:

• resolve_link
• list_bases
• discover_tables
• find_table
• list_tables
• get_table
• get_record
• query_records
• create_records
• update_records
• upsert_records
• delete_records
• create_table
• rename_table
• delete_table
• create_field (including linked fields via link.table_id / link.table_name, plus duplex links via link.back_field_name)
• rename_field
• update_field
• delete_field
• list_folder

Best-use guidance#

• Prefer direct Base links or explicit identifiers ( app_token, table_id) when available.
• Discovery helpers are convenience features; direct link/token workflows are more reliable.
• Use discover_tables for broad, self-discoverable scans when you do not know the app token.
• Inspect schema before writing when field names are uncertain.
• Query/list before update when locating existing records.
• Avoid hardcoding a fixed app/table in workflow logic; derive the target from the current user request (link, base, table name, or explicit IDs).

Workflow: add customer record (dynamic, non-hardcoded)#

Use this when user asks to “add customer”.

1. Resolve target table first.
   • If user provides a Base link, run resolve_link and use returned app_token + table_id.
   • If no table is specified, run list_tables / find_table and choose best match ( Customers, Customer, 客户).
   • If multiple likely matches exist, ask one concise disambiguation question.
2. Inspect schema before write.
   • Run get_table and read writable fields.
   • Map by field names (for example: Customer Name, Company Name, City, State, Country, Phone, Email, Status).
3. Build minimal safe payload.
   • Fill only fields provided by user.
   • If user says location only (e.g., “Jiji at Penang”), map Customer Name=Jiji, City/State=Penang when present.
   • For select fields, only use valid options; otherwise ask or write to a text fallback field.
4. Create and verify.
   • Run create_records with one record first.
   • Confirm success by returned record_id (and optionally get_record / query_records if needed).
   • Report exact fields written.
5. Handle blockers explicitly.
   • If FEISHU_NOT_CONFIGURED or permission errors occur, stop and ask user to reconnect/authorize Feishu; do not loop retries.

Safety#

Destructive operations are supported but should be disabled by default unless explicitly needed.

Config flag:

• allowDelete: false (recommended default)

When deletion is disabled, destructive actions should be blocked:

• delete_records
• delete_field
• delete_table

Requirements#

• OpenClaw with Feishu channel configured
• valid Feishu credentials already present in OpenClaw config
• access to the target Base/Bitable resources

Notes#

• This plugin works best as a link-first / token-first Feishu Base tool.
• Some discovery flows depend on what Feishu APIs expose in the current tenant/account context.
• For field creation and schema mutation, prefer serialized one-by-one writes. Parallel field creation can hit Feishu-side limits and cause intermittent 400 failures.
• Credential resolution prefers explicit account_id, then active runtime/session account context when available, then runtime-injected channels.feishu, and finally persisted OpenClaw config from OPENCLAW_CONFIG_PATH or ~/.openclaw/openclaw.json when some runtime paths do not inject Feishu config consistently.