Skip to main content
Glama

Perisclaw: Use ChatGPT and Claude with your WhatsApp

Server Details

Connect your own WhatsApp number to Cursor, Claude and ChatGPT: catch up, search, send replies.

Ownership verified
Status
Healthy
Uptime
100.0% over the last 24 hours
OAuth
Works in Glama
Last Tested
Transport
Streamable HTTP · MCP 2025-11-25
URL
Repository
periskope-org/perisclaw-plugin
GitHub Stars
0
ChatGPT
–
Claude
–

TDQS

A3.6/5.0

Scored across 18 tools

Disambiguation3/5

Most tools have clear distinct targets, but the card-display pair workspace_open and workspace_conversation overlaps in mechanism (both render an interactive card) and the boundary only becomes clear after reading long descriptions. Contact/chat resolution is spread across four similar tools (search_contacts, contacts_get_batch, contact_get, search_chats), and workspace_conversation references reply_prepare/reply_send which do not exist in this set, adding confusion.

Naming Consistency3/5

Everything is snake_case, which helps, but verb placement is inconsistent: noun-first suffixed verbs (automation_get, chat_settings_get, contact_get, media_download) sit alongside verb-first tools (get_messages, get_profile, search_chats, search_messages). The workspace_* and *_recent/*_status names follow yet another pattern.

Tool Count4/5

18 tools is slightly heavy but defensible for a WhatsApp client spanning chats, messages, contacts, media, automations and UI cards. Each grouping (read automation, read chat settings, resolve contacts) earns its place, though the four separate contact/chat lookup tools could be consolidated.

Completeness3/5

The surface is heavily read-only: automations and chat_settings can be listed and fetched but never created or updated, and there is no direct send/mark-read operation, with replies only possible through workspace cards. The dangling references to reply_prepare/reply_send suggest removed or missing tools, and no AI/ChatGPT/Claude interaction is exposed despite the server name.

Available Tools

18 tools
automation_getGet AutomationA
Read-only
Inspect

Read one Automation by its canonical id. A durable flow includes health over the last 7 days: runs, succeeded/no_action/failed, failure_streak, model_calls, cost_usd, cost_per_run_usd, calls_per_run and last_error.

Inputs: id (required).

automation_listList AutomationsA
Read-only
Inspect

List every Automation, regardless of its internal execution engine, including notifications, recurring tasks, scheduled behavior, event rules, and durable flows. If it can fire in the future, it appears here. Finished one-time Automations (off, with status completed or cancelled) are left out and counted in finished_hidden; pass include_finished to list them, e.g. to check whether a past reminder went out. enabled/watch/scheduled means armed only; last_run_at, latest_run, and active_run_count are the evidence of whether it has ever started and what happened. Each durable flow's health gives its last 7 days of outcomes, failure_streak and recorded model cost. Rows are summaries; get one Automation by id for its definition, schedule, full health and latest run.

Inputs: include_finished.

automation_runsList Automation RunsB
Read-only
Inspect

Inspect automation runs and durable progress state. Each run carries its failure reason (last_error_code/last_error_message) and usage: the model calls, tokens and cost_usd it paid for (null when none were recorded).

Inputs: limit, since, status, active_only, +1 more.

chat_settings_getGet Chat SettingsA
Read-only
Inspect

Read the chat_settings row for a chat. Returns null if no row exists (meaning the chat uses defaults — mode=0, no per-chat notes). This is READ-ONLY and never starts durable work. Finite follow-through belongs to a Mission; standing monitoring or replies belong to an Automation.

Inputs: chat_id (required).

chat_settings_listList Chat SettingsA
Read-only
Inspect

List all chats with chat_settings rows. Useful for auditing which chats have explicit notes. Chats without rows use defaults and don't appear here.

Inputs: no named parameters.

chats_recentList Recent ChatsA
Read-only
Inspect

List the operator's WhatsApp chats, MOST RECENT ACTIVITY FIRST — each with chat_id, name, unread_count, last-activity time, and labels. Optionally restrict to chats with unread messages, and cap the count.

Inputs: limit, unread_only.

contact_getGet ContactA
Read-only
Inspect

Read a contact + everything the agent already knows about them (metadata.enrichment). Call this before answering questions about a person or deciding if a newly-observed fact is novel.

Inputs: contact_id (required), include_enrichment.

contacts_get_batchGet ContactsA
Read-only
Inspect

Resolve a list of JIDs to {name, pushname, lid} in one round trip. Use when chat context has bare JIDs without paired name labels.

Inputs: contact_ids (required).

get_group_inviteGet Group InviteB
Read-only
Inspect

Get the invite URL for a group (only works if operator is admin).

Inputs: chat_id (required).

get_messagesGet MessagesA
Read-only
Inspect

Read recent messages from ONE chat, MOST RECENT FIRST. Returns sender (JID), canonical sender_identity / @sender_mention (including for LID group senders), from_me, body, type, attachment filename/mimetype, timestamp, reactions, and any quoted message. Messages deleted for everyone read "This message was deleted" with type revoked and their original text in prev_body. Empty / system / undecryptable rows are skipped, so this is the clean way to read or summarise a conversation. Page to older messages with before. pinned lists what is pinned in the chat now (absent when nothing is).

Inputs: chat_id (required), limit, before.

get_profileConnected accountA
Read-only
Inspect

Use this when you need to know which WhatsApp account this connection uses, for example when the user has several Perisclaw connections. Returns a stable account id and the phone number.

Inputs: no named parameters. Returns: id, name, nickname

media_downloadDownload MediaA
Read-only
Inspect

Fetch the photo, voice note, video or document attached to a message in the user's own WhatsApp so it can be viewed or described here. Use it only for a message the user is asking about. Pass that message's message_id. Files up to 10 MB are shown here; larger files can't be opened in this chat.

Inputs: message_id (required).

reply_statusCheck replyA
Read-only
Inspect

Use this when a reply was not confirmed, to see whether it went out. Returns reply: null when nothing was saved. Never sends anything.

Inputs: card_id, draft_id. Returns: reply

search_chatsSearch ChatsA
Read-only
Inspect

Fuzzy-match a chat (group or DM) by name. Same matching as search_contacts.

Inputs: query (required), limit.

search_contactsSearch ContactsA
Read-only
Inspect

Fuzzy-match a contact by name, or find the contact an exact email address belongs to (its contact_id is also their WhatsApp chat). Handles typos, phonetic variants, and word-order swaps.

Inputs: query (required), limit.

search_messagesSearch MessagesA
Read-only
Inspect

Hybrid semantic and full-text search across WhatsApp messages. Returns relevant messages with authoritative chat and sender metadata, authorship provenance, and quoted-message context when present. Use this for finding past conversations by meaning or keyword. Pass a natural-language query. The semantic path uses the full text; the FTS path safely tokenises it as ANY-of-these words. FTS operators, quotes, and * are stripped rather than interpreted.

Inputs: query (required), limit, since, until, +3 more.

workspace_conversationShow reply cardA
Read-only
Inspect

Use this when the user wants to reply to a WhatsApp chat. Shows a card with the latest messages and your draft; the user edits it and presses Send in the card, which sends it. Never call reply_prepare or reply_send for a draft shown in a card: if the user says "send it", tell them to press Send in the card. To change the draft, call this again with new draft_text. Resolve the chat first. Sends nothing by itself.

Inputs: chat_id (required), draft_text, quote_message_id. Returns: chat, quoted, card_id, profile, +5 more

workspace_openShow catch-upA
Read-only
Inspect

Use this to answer whenever the user asks what needs their attention or reply on WhatsApp, or to catch up. First read the relevant chats with chats_recent and get_messages, then always show this card instead of writing the list out: pass up to 8 chats in priority order, each with a one-line reason naming who is waiting and on what ("Sarah asked you to confirm Friday"). The card is the answer, so add at most one or two sentences. Without chats it shows the most recent chats; that is only for opening Perisclaw from the sidebar. Does not mark anything read.

Inputs: chats, unread_only. Returns: chats, profile, skipped, read_only, +1 more

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      WA MCP is a hosted WhatsApp MCP server. It connects your own WhatsApp number to Claude, Claude Code, ChatGPT, Grok Bot, Cursor or any MCP client. Free plan with no time limit, nothing to install.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      A personal MCP for WhatsApp, with voice-note and call transcripts. Search and summarize your chats, read what was said in voice notes and calls, and organize chats, labels, notes, reminders and scheduled messages. Runs locally with the Kaption extension for WhatsApp Web (Chrome, Edge) or the Kaption desktop app. No send-now tool. Setup: https://kaptionai.com/mcp/
      18
      1,058 npm
      1
      Business Source 1.1
    • A
      license
      B
      quality
      B
      maintenance
      Connects Claude, Cursor and other AI assistants to a WhatsApp Business API workspace so they can search contacts and tags, read inbox conversations, check template approval and message delivery status, review broadcast and CRM deal data, and send single opt-in messages or create contacts and deals. Bulk broadcasts, drips and deletes are deliberately excluded, keeping assistants limited to safe, individual actions.
      17
      245 npm
      MIT
    Try in Browser

    Glama MCP Gateway

    Add one secure layer between your agents and this server.