LeadCeleris
Model Context Protocol

Operate LeadCeleris from Codex or Claude

LeadCeleris MCP gives a trusted AI client a controlled, workspace-scoped set of tools. Ask questions about your CRM, review conversations, maintain knowledge, and perform approved actions without building a custom integration.

Create your API key Read setup steps

LeadCeleris Workspace

Remote MCP server online

11 tools

Your AI client

Codex or Claude Code

Revocable access

One workspace per key

Useful actions

CRM, inbox, pipeline, knowledge

What MCP changes

Talk to your workspace instead of clicking through every screen

MCP is a standard connection between an AI client and a set of tools. The model does not receive your password. LeadCeleris validates a revocable API key, limits every request to its workspace, and exposes clearly labeled read and write actions.

Daily lead briefing

Ask for qualified leads, stale conversations, or contacts with high scores and turn the result into a prioritized call list.

Try: Review my workspace. Show the highest-priority leads that need a human response, with a one-line reason for each.

Conversation copilot

Summarize a long WhatsApp history, update internal notes, and prepare or send the next response after you approve it.

Try: Summarize this contact's conversation, identify the unresolved question, and draft a concise reply. Do not send it yet.

Knowledge maintenance

Turn product notes, service details, prices, policies, and FAQs into structured knowledge the sales AI can use.

Try: Add a service called Express Installation with this description and price range, then confirm what was saved.

Pipeline operations

Review new orders, bookings, and escalations together, then identify anything blocked or missing an owner.

Try: List recent orders, bookings, and escalations. Flag anything that appears unresolved or needs human attention.

Agent improvement

Read the current prompt and escalation rules, propose a focused change, and apply it only after review.

Try: Review the current escalation rules. Suggest a rule for serious enterprise buyers, but ask before changing the configuration.

CRM cleanup

Find unqualified or misclassified contacts, normalize notes, and keep AI enabled only where it makes sense.

Try: Find contacts marked as leads whose recent messages look like spam. Show your reasoning and ask before updating them.

One shared prerequisite

Create a workspace key once

Sign in to LeadCeleris, open Developers, enter a name such as “My Codex” or “My Claude,” then select Create key. Copy the full key immediately. It is shown only once.

01

Create the key

Open App → Developers → API Keys. Use a separate key for every AI client.

02

Save it securely

Treat the lc_live_… value like a password. Never put it in a public repository.

03

Revoke when needed

Remove the key from Developers to disconnect that client immediately.

Guide 1

Connect OpenAI Codex

  1. 1. Set LEADCELERIS_API_KEY in the environment that launches Codex.
  2. 2. Open ~/.codex/config.toml, or use a trusted project’s .codex/config.toml.
  3. 3. Add the configuration below, save, and restart Codex.
  4. 4. Open MCP status in the Codex app, IDE, or CLI and confirm leadceleris is connected.

Windows PowerShell, save the key for your user:

[Environment]::SetEnvironmentVariable(
  "LEADCELERIS_API_KEY",
  "PASTE_YOUR_LEADCELERIS_KEY",
  "User"
)

Completely close Codex after setting the variable. Then add this server configuration and reopen Codex:

[mcp_servers.leadceleris]
url = "https://app.leadceleris.com/api/mcp"
bearer_token_env_var = "LEADCELERIS_API_KEY"
default_tools_approval_mode = "writes"

The writes approval mode keeps read-only research convenient while asking before tools that modify contacts, send messages, change the agent, or add knowledge.

Full Codex walkthrough Official Codex MCP docs

Guide 2

Connect Anthropic Claude Code

  1. 1. Install and sign in to Claude Code, then open your terminal.
  2. 2. Replace the placeholder in the command below with the one-time LeadCeleris key.
  3. 3. Run the command, start Claude Code, and enter /mcp.
  4. 4. Confirm the server is connected, then ask for a workspace overview.
claude mcp add --transport http --scope user leadceleris \
  https://app.leadceleris.com/api/mcp \
  --header "Authorization: Bearer YOUR_LEADCELERIS_API_KEY"

This guide is for Claude Code. Claude.ai hosted custom connectors use an OAuth connection flow. LeadCeleris currently supports revocable bearer keys, so direct Claude.ai setup is not advertised yet.

Full Claude walkthrough Official Claude Code MCP docs

Available tools

Useful access, with visible boundaries

Read tools inspect the workspace. Write tools can change data or contact a customer and should stay approval-gated in your AI client.

Workspace overview

Read

Channels, CRM totals, open escalations, orders, bookings, and AI status.

List contacts

Read

Search contacts and filter lead stages, with scores, sources, intent, and recent activity.

Inspect a contact

Read

Read one contact's CRM details, human-only labels, and recent conversation messages.

Update contacts

Write

Change a contact name, lead stage, notes, or whether the sales AI is enabled.

List conversations

Read

Find recent conversations and their latest message.

Send a message

Write

Send a WhatsApp text inside an existing contact conversation.

Review the pipeline

Read

Inspect recent orders, bookings, and human escalations.

Read agent configuration

Read

Review the advanced prompt, escalation rules, and configured GET tools.

Update agent configuration

Write

Change the advanced prompt or natural-language escalation rules.

List knowledge

Read

Read active products, services, FAQs, policies, promotions, and locations.

Create knowledge

Write

Add a product, service, FAQ, policy, promotion, location, or other entry.

Start with a safe first prompt

“Connect to LeadCeleris, show me a workspace overview, and list the five most recent conversations. Read only. Do not send messages or change any data.”

Open Developers