https://api.mcp.modjo.ai/v1/mcp

{   "mcpServers": {     "modjo": {       "type": "http",       "url": "https://api.mcp.modjo.ai/v1/mcp",       "headers": {         "Authorization": "Bearer votre-clé-api-modjo-ici"       }     }   } }

---  ## Overview  This skill enables the agent to use the **Modjo MCP server** to retrieve and analyze sales data: accounts, contacts, deals, calls, emails, call transcripts, and AI-powered insights via Modjo's built-in and custom agents.  The agent should **always read this skill before making any Modjo MCP tool call** to follow the correct tool chaining patterns and avoid common pitfalls.  ---  ## Tool reference (accurate schemas)  Filters are passed inside a `filters` object. Pagination uses `pagination: { cursor, limit }` (1–50, default 10).  ### Data retrieval tools  | Tool | Required filters | Optional filters | Key returns | | --- | --- | --- | --- | | `get_accounts` | `name` (string, required) | — | `crmId`, `name`, `crmLink`, `crm` | | `get_contacts` | — | `name` (string) | `crmPersonId`, `name`, `email`, `phoneNumber`, `jobTitle`, `crmLink` | | `get_deals` | — | `name`, `account.crmIds[]`, `status[]`, `closeDate{start,end}`, `amount{min,max}`, `source[]`, `startDate{start,end}`, `lossReason` | `crmId`, `accountCrmId`, `name`, `status`, `amount`, `startDate`, `closeDate`, `lossReason`, `source` | | `get_calls` | — | `name`, `accounts.crmIds[]`, `contact.crmIds[]`, `deal.crmIds[]`, `dateRange{start,end}`, `userIds[]` | `id`, `name`, `status`, `date`, `duration`, `direction`, `phoneProvider`, `crmLink`, `dealId`, `summary` | | `get_emails` | — | `subject`, `accounts.crmIds[]`, `contact.crmIds[]`, `deal.crmIds[]`, `dateRange{start,end}`, `userIds[]` | `id`, `subject`, `date`, `from`, `to`, `dealId`, `accountId` | | `get_transcript` | `callIds[]` (number[]) | — | Transcript blocks: `startTime`, `endTime`, `content`, `speakerName` | | `get_users` | — | `name`, `email` | `id`, `email`, `name`, `role`, `jobTitle`, `jobDepartment` | | `get_agents` | — | `search` (string), `creatorOrigin` ("modjo" or "user") | `uuid`, `name`, `description` |  ### AI analysis tools  | Tool | Required params | Optional params | Returns | | --- | --- | --- | --- | | `ask_anything_on_call` | `id` (string — call ID), `question` (string) | `agentUuid`, `language` | AI-generated answer about the call | | `ask_anything_on_account` | `crmId` (string — account CRM ID), `question` (string) | `agentUuid`, `language` | AI-generated answer about the account | | `ask_anything_on_deal` | `crmId` (string — deal CRM ID), `question` (string) | `agentUuid`, `language` | AI-generated answer about the deal |  ---  ## Critical schema notes (common pitfalls)  1. **`get_accounts` requires `name`** — the `filters.name` field is required, not optional. You cannot fetch all accounts without a name filter. If the user says "list all accounts", try a very short common substring or inform them that a name filter is needed. 2. **`get_calls` and `get_emails` use nested filter objects** — account, contact, and deal filters are nested:     - Correct: `filters: { accounts: { crmIds: ["001..."] } }`     - Wrong: `filters: { accountCrmIds: ["001..."] }` 3. **`dateRange` format** — uses `{ start: "YYYY-MM-DD", end: "YYYY-MM-DD" }` (ISO date strings, not datetimes). 4. **`get_transcript` takes `callIds` as numbers** — not strings. The call `id` from `get_calls` is a number (e.g., `74969`), pass it directly: `filters: { callIds: [74969] }`. 5. **`ask_anything_on_call` takes `id` as a string** — despite the call ID being a number in `get_calls`, the `ask_anything_on_call` tool expects it as a string: `id: "74969"`. 6. **`get_deals` returns `accountCrmId`** — this field links the deal to its parent account. Use it to cross-reference with `get_accounts` results. 7. **`get_deals` filtering capabilities** — `get_deals` supports rich server-side filtering. Use these instead of fetching all deals and filtering client-side:     - `status` — string array. Valid values: `"Open"`, `"Closed won"`, `"Closed lost"`, `"Closed"`, `"Deleted"`.     - `closeDate` — dateRange object: `filters: { closeDate: { start: "2026-03-23", end: "2026-04-05" } }`     - `amount` — range object (both fields optional): `filters: { amount: { min: 10000, max: 50000 } }`     - `source` — string array (max 50): `filters: { source: ["Inbound", "Outbound"] }`     - `startDate` — dateRange object, same format as `closeDate`     - `lossReason` — string, case-insensitive partial match (max 255 chars)     - All filters can be combined. 8. **`get_calls` returns `summary`** — most calls already have a pre-generated summary. Check this before calling `ask_anything_on_call` for basic questions — the summary may already answer them. 9. **Pagination** — all list tools return `pagination: { hasNextPage, nextCursor, limit }`. If `hasNextPage` is true, pass `nextCursor` as `cursor` in the next call. Max 50 per page. Always use `limit: 50` when you need comprehensive results. 10. **`get_contacts` does NOT require `name`** — unlike `get_accounts`, the name filter is optional. 11. **`userIds` in `get_calls`/`get_emails`** — these are number arrays, not strings. Get the numeric `id` from `get_users` first.  ---  ## Using the `ask_anything_on_*` tools  These are Modjo's most powerful capabilities — AI agents that analyze calls, accounts, and deals.  **When to use them vs raw data:**  - **Prefer `ask_anything_on_call`** over `get_transcript` for qualitative insight (objections, sentiment, competitors, next steps, coaching feedback). - **Use `get_transcript`** only when verbatim quotes are needed. - **Prefer `ask_anything_on_account`** for holistic relationship summaries — saves chaining multiple calls/emails lookups. - **Prefer `ask_anything_on_deal`** for deal health, risk, and next steps in one query.  **How to pick an agent (`agentUuid`):**  1. Call `get_agents` first when the user needs specialized analysis (coaching, competitor analysis, MEDDPICC, churn detection, etc.) 2. Filter by `creatorOrigin: "modjo"` for built-in agents, `creatorOrigin: "user"` for custom agents 3. If no specific agent fits, **omit `agentUuid`** — the default agent handles general questions well 4. For French-speaking users/data, set `language: "fr"` to get responses in French  **Built-in Modjo agents (always available):**  | Agent | UUID | Best for | | --- | --- | --- | | CallQualifier | `3a7753e1-...` | Prospect profiles, qualification gaps | | CallSummary | `741e9ffc-...` | Quick call recaps, CRM-ready | | DealBriefing | `1204e84f-...` | Exec-ready deal summaries | | EmailFollowUp | `b2a9ae1b-...` | Auto-draft follow-up emails | | MeetingPrepper | `c0a76fd7-...` | Pre-meeting briefs | | NextStepper | `09715241-...` | Action items and commitments |  ---  ## Playbooks  ### 1. 🔎 Call lookup for a specific company  **Trigger**: "Show me calls with [Company]", "What was discussed with [Company]?", "Any recent calls about [topic]?"  **Steps**:  1. `get_accounts` → filter by company name → extract `crmId` 2. `get_calls` → filter by `accounts.crmIds` → list calls (you can filter directly by account, you don't always need to go through deals first) 3. *(Optional)* `get_deals` → filter by `account.crmIds` → if deal-level filtering is needed 4. For interesting calls → check the `summary` field first, then use `ask_anything_on_call` for deeper analysis  **Output**: Summary of calls (date, duration, direction, key points), with AI insights on notable calls.  ### 2. 📋 Prepare a meeting  **Trigger**: "Help me prepare for my meeting with [Person/Company]", "Brief me before my call"  **Steps**:  1. `get_contacts` → filter by contact name → extract `crmPersonId`, `jobTitle` 2. `get_accounts` → filter by company name → extract `crmId` 3. `get_deals` → filter by `account.crmIds` → pipeline context 4. `get_calls` → filter by `contact.crmIds` or `accounts.crmIds` → recent interactions (use `limit: 50` + `dateRange` for last 90 days) 5. `get_emails` → filter by `contact.crmIds` or `accounts.crmIds` → recent email exchanges 6. `ask_anything_on_account` → "What are the key topics discussed, open issues, and relationship status with this account?" 7. *(Optional)* `ask_anything_on_deal` → "What are the current risks and recommended next steps?"  **Output**: Structured meeting brief — contact profile, account overview, deal status, recent interaction summary, AI-generated talking points and risks.  ### 3. 🔥 Deal prioritization  **Trigger**: "Which deals should I focus on?", "What are my hot deals?", "Prioritize my pipeline"  **Steps**:  1. `get_deals` → filter by `status: ["Open"]` and optionally `closeDate` for the relevant window and/or `amount: { min: X }` to focus on material deals → use `limit: 50` 2. `get_calls` → filter by `deal.crmIds` → assess activity recency and frequency 3. `get_emails` → filter by `deal.crmIds` → assess email engagement 4. `ask_anything_on_deal` → for top deals, ask about momentum, blockers, and buying signals  **Scoring framework**:  - 🔴 Hot: Recent interactions + high amount + close date imminent + positive AI signals - 🟡 Warm: Some activity + medium amount + close date within quarter - 🔵 Cold: No recent interactions + stalled + past close date + low engagement  **Output**: Ranked deal list with priority, AI rationale, and recommended next action per deal.  ### 3b. 📅 Pipeline review (weekly/monthly)  **Trigger**: "Pipeline review for this week", "Deals closing this month", "What's overdue?"  **Steps**:  1. `get_deals` → filter by `status: ["Open"]` + `closeDate: { start: "<period_start>", end: "<period_end>" }` → deals expected to close in the window 2. `get_deals` → filter by `status: ["Open"]` + `closeDate: { start: "<past_date>", end: "<yesterday>" }` → overdue deals 3. `get_deals` → filter by `status: ["Closed won", "Closed lost"]` + `closeDate` for same window → recently closed 4. For each deal → `get_calls` + `get_emails` filtered by `deal.crmIds` → assess recent activity 5. `ask_anything_on_deal` → for key deals, ask about health, risks, and next steps  **Output**: Structured pipeline review — overdue deals, deals closing soon, recently won/lost, with activity levels and AI-driven recommendations per deal.  ### 4. 📊 Win/Loss analysis  **Trigger**: "Why did we lose deals?", "Win patterns?", "Analyze closed deals in [period]"  **Steps**:  1. `get_deals` → filter by `status: ["Closed won"]` + optionally `closeDate` for the period → won deals 2. `get_deals` → filter by `status: ["Closed lost"]` + same `closeDate` → lost deals. Use `lossReason` filter for a specific loss pattern 3. `get_calls` → filter by `deal.crmIds` → associated calls 4. `get_agents` → search for a win/loss or analysis agent 5. `ask_anything_on_deal` → for sample of won/lost deals, ask about key outcome factors, competitor mentions, pricing objections  **Output**: Win/loss report — quantitative breakdown (win rate, deal size, top loss reasons) + AI qualitative insights + recommendations.  ### 5. 👤 Contact intelligence  **Trigger**: "Tell me about [Contact]", "What's my history with [Person]?"  **Steps**:  1. `get_contacts` → filter by name → extract `crmPersonId`, `jobTitle`, `email` 2. `get_calls` → filter by `contact.crmIds` → interaction history 3. `get_emails` → filter by `contact.crmIds` → email threads 4. `ask_anything_on_call` → for recent calls, assess tone, concerns, and sentiment  **Output**: Contact profile + interaction timeline + AI relationship insights.  ### 6. 📅 Activity review (team or individual)  **Trigger**: "What happened this week?", "Show me [Rep]'s activity", "Team activity in Q1?"  **Steps**:  1. *(If filtering by rep)* `get_users` → filter by name/email → extract numeric `id` 2. `get_calls` → filter by `dateRange` and `userIds` → retrieve calls 3. `get_emails` → filter by `dateRange` and `userIds` → retrieve emails 4. *(Optional)* Enrich with `get_accounts`/`get_deals` for context  **Output**: Activity summary — call count, email count, deals touched, highlights by rep or team.  ### 7. 🏢 Account deep dive  **Trigger**: "Full picture of [Company]", "Status of our relationship with [Account]?"  **Steps**:  1. `get_accounts` → filter by name → extract `crmId` 2. `ask_anything_on_account` → comprehensive AI summary (this alone provides great insight) 3. *(Optional)* `get_deals` → filter by `account.crmIds` → pipeline detail 4. *(Optional)* `ask_anything_on_deal` → per active deal, assess status and next steps  **Output**: Account brief — relationship history, key themes, active deal status, recommended actions.  ### 8. 🤖 Agent-assisted custom analysis  **Trigger**: When the user asks for coaching feedback, competitor analysis, objection handling, MEDDPICC, BANT, churn detection, or any specialized analysis type.  **Steps**:  1. `get_agents` → search by keyword (e.g., "coaching", "competitor", "MEDDPICC", "churn") → extract `uuid` 2. Locate the target entity via `get_calls`, `get_accounts`, or `get_deals` 3. Use the agent's `uuid` in the appropriate `ask_anything_on_*` tool  **Notable custom agents available**: CallScorer, ChurnInspector, Deal Challenger (MEDDPICC), DealRiskDetector, ProductFeedbacksAggregator, QuickBANT, SalesCoach, MeddicValidator, Objectionshunter.  **Output**: Specialized analysis powered by the matched Modjo AI agent.  ---  ## General principles  - **Always chain tools** — never rely on a single tool when richer context is available from combining multiple. - **Start broad, then narrow** — resolve account/contact CRM IDs first, then filter calls/emails by those IDs. - **Check summaries first** — `get_calls` returns a `summary` field. Read it before making additional `ask_anything_on_call` calls. - **Prefer `ask_anything_on_*` for qualitative analysis** — these provide richer, AI-synthesized insights than raw transcript reading. - **Use `get_agents` proactively** — when a specialized analysis type is requested, check for a matching agent before defaulting to generic questions. - **Use `get_users` for team filtering** — resolve rep names to numeric `id` values before filtering calls/emails by `userIds`. - **Respect pagination** — max 50 records per page. Use `limit: 50` and paginate with `nextCursor` when needed. - **Synthesize, don't dump** — always summarize and interpret data, never raw-dump JSON results. - **Surface next actions** — conclude with clear, actionable next steps for the sales rep. - **Language** — if the user writes in French or the data is in French, set `language: "fr"` on `ask_anything_on_*` calls.

curl -H "Authorization: Bearer votre-clé" https://api.mcp.modjo.ai/v1/mcp