Tuesday, August 4, 2026
The Strama MCP Server: Run Your Outbound Motion From Claude
Strama's MCP server connects your Strama workspace to Claude and other MCP clients. Once connected, you can run your whole outbound motion in conversation: search your CRM, find net-new prospects, qualify them against a scorecard, build a campaign, generate personalized sequences, and send email and LinkedIn messages — without leaving the chat.
This page covers what the connector does, how to set it up, prompts to try, the full tool reference, and how the server handles data. If you are an IT administrator evaluating Strama for your organization, skip to Security and data handling.
What you can do with it
The connector exposes 20+ tools spanning the full outbound workflow. In practice they compose into four kinds of work:
- Ask questions about your data. Search contacts, companies, campaigns, sequences, deals, meetings, tasks, notes, and the email and LinkedIn inbox with free-text or structured filters.
- Build your list. Find net-new prospects in the Strama lead database or through your own LinkedIn Sales Navigator seat, import specific records from HubSpot or Salesforce, and enrich missing emails, phone numbers, and LinkedIn profiles.
- Qualify. Define an A–F scorecard with your exact criteria, then grade contacts and companies against it and read back per-criterion verdicts.
- Run outreach. Create a campaign, enroll contacts, generate AI sequences in your sender's voice, and start, pause, or complete them — or send a single email or LinkedIn message directly.
Setup
The Strama MCP server is a remote server using OAuth 2.0. There is nothing to install and no API key to paste or rotate.
- Make sure you have a Strama account and belong to at least one team. If you do not have one, sign up at strama.ai.
- In Claude, go to Settings → Connectors and add the Strama connector. If you are adding it as a custom connector, the server URL is
https://mcp.strama.ai/mcp. - Click Connect. You will be redirected to Strama to sign in and authorize access.
- Approve the requested access. You are returned to Claude and the tools become available.
The connector acts as you. It can only see and change the data your own Strama login can see, in the teams you belong to. Every tool accepts an optional teamId; omit it and the tool uses your default team. Ask Claude to "list my Strama teams" to see them.
You can revoke access at any time from Claude's connector settings or from your Strama account settings.
Prompts to try
Each of these exercises a different part of the server. They are written the way you would actually type them.
1. Read your pipeline
Which companies are graded A or B on my "Mid-Market RevOps" scorecard? For each one, list the contacts we have and whether they are enrolled in any campaign.
Expected: Claude calls search against scorecard runs, companies, contacts, and sequences, then returns a grouped list with grades and enrollment status. Read-only — nothing changes, and no credits are consumed.
2. Find and add net-new prospects
Find VP Sales and CRO titles at software companies in the United States with 51–200 employees that raised in the last year. Show me the top 20, then add the ones at companies we have never contacted.
Expected: Claude calls discover_prospects, shows the results with names, titles, and companies, and flags any that already exist as contacts. After you pick, it calls add_prospects_from_search to create them. The search results are saved server-side for about 30 days, so you can come back and add more later without re-running it. Lead-database searches consume credits; Sales Navigator searches through your own seat do not.
3. Build a rubric and qualify accounts
Create a Company scorecard called "Series B SaaS ICP" with criteria for headcount growth, an existing outbound team, and a US-based go-to-market org. Then grade the 20 companies I imported last week against it.
Expected: Claude calls create_scorecard with your criteria stored verbatim — nothing is AI-reworded — then grade_entities. Grading is a background job, so Claude returns a task id and polls get_task_result until each company has a grade, a verdict summary, and per-criterion reasoning.
4. Stand up a campaign without sending anything
Create a campaign called "Q3 RevOps Outbound" using the Series B SaaS ICP scorecard for companies and my RevOps Leaders scorecard for contacts. Enroll every A-graded contact, generate their sequences, and leave everything idle — do not start sending.
Expected: Claude calls create_campaign, enroll_entities, and generate_sequences. You get the drafted steps back to review. Nothing sends until you explicitly ask to start the sequences, which is a separate tool call that requires your approval.
5. Pull from your CRM and fill the gaps
Import HubSpot company 12345 and its contacts into Strama, then enrich any contacts missing an email address.
Expected: Claude calls import_crm_records with includeContacts, so the account and its people land in Strama pre-linked to their CRM records and will not duplicate on sync. It then calls enrich_contacts for the missing emails and reports which lookups succeeded. Records already imported come back as already linked rather than duplicated, so re-running is safe.
Tool reference
Every tool is annotated so Claude knows whether it reads or writes. Read-only tools can run without a per-call confirmation. Tools that write, spend credits, or send messages always prompt you first.
Read-only
| Tool | What it does |
|---|---|
search | Query any Strama object — contacts, companies, campaigns, senders, sequences, campaign companies, scorecards, scorecard runs, meetings, deals, tasks, leads, notes, email messages, LinkedIn messages, and Scout personas — with free-text and structured filters. |
describe | Returns the filterable and writable properties for each object type, so Claude builds valid queries instead of guessing field names. |
list_teams | Lists the Strama teams you belong to, with your role in each. |
get_task_status | Checks progress on a background job. |
get_task_result | Fetches the output of a completed background job. |
Building your list
| Tool | What it does |
|---|---|
discover_prospects | Searches for net-new people beyond your CRM, from the Strama lead database or your connected LinkedIn Sales Navigator seat. Creates no records on its own. |
add_prospects_from_search | Turns selected results from a prior discovery search into contacts, optionally enrolling or grading them in the same step. |
save_scout_persona | Saves a search's lead-level filters as a reusable persona. |
import_crm_records | Imports specific HubSpot or Salesforce records by id, pre-linked so CRM sync will not duplicate them. |
enrich_contacts | Looks up a contact's missing email, phone, current company and title, or LinkedIn profile. Consumes credits. |
enrich_companies | Resolves a company's LinkedIn page from its name and website domain. Consumes credits, refunded when nothing is found. |
upsert | Creates or updates contacts, companies, meetings, deals, tasks, leads, and notes. Only the fields you pass are changed. |
Qualification
| Tool | What it does |
|---|---|
create_scorecard | Creates a reusable A–F rubric from criteria you supply, stored exactly as written. Grades nothing by itself. |
grade_entities | Researches and grades contacts or companies against a scorecard, returning a grade, verdict, and per-criterion results. Consumes credits. |
Outreach
| Tool | What it does |
|---|---|
create_campaign | Creates a campaign with an outreach plan and links the contact and company scorecards that qualify it. Enrolls no one and sends nothing. |
enroll_entities | Adds existing contacts or companies to a campaign. Contact enrollments start idle. |
generate_sequences | Writes personalized outreach steps for enrolled contacts in your sender's voice. Sends nothing unless you explicitly ask it to auto-start. |
set_sequence_state | Starts, pauses, or completes sequences. Setting a sequence to Running begins sending real outreach on its schedule. |
remove_sequence_channel | Destructive. Permanently deletes all steps of one channel from a sequence; the generated copy cannot be recovered. |
send_email | Sends a real email immediately, either as a new thread or a reply on an existing one. |
send_linkedin_message | Sends a real LinkedIn message immediately through your connected LinkedIn account. |
What sends real messages
Three tools put messages in front of real people, and it is worth being explicit about them:
send_emailandsend_linkedin_messagesend immediately. There is no draft or approval step inside Strama — Claude's own confirmation prompt is the approval step.set_sequence_statewith a target of Running starts automated outreach on each sequence's schedule.
Everything upstream of those three — discovering, importing, enriching, grading, creating campaigns, enrolling, generating copy — changes data in Strama but sends nothing. If you want to work with the connector without any risk of outbound going out, you can stay entirely within those tools and review the generated sequences in the Strama UI before starting them.
Two guardrails apply regardless of how a send is triggered: sends are blocked to contacts who have opted out, and they are blocked once a sender's daily limit is reached.
Known limitations
- Long-running work is asynchronous. Enrichment, grading, and sequence generation return a task id and finish in the background, typically in one to five minutes. Claude polls for the result; on a slow job you may need to ask it to check again.
- Discovery consumes credits. Strama lead database searches cost one credit per page of roughly 20 results, and a single search session can page through at most 250 prospects. Sales Navigator searches through your own seat are free but require a connected seat with Sales Navigator access.
- Scout personas save lead-level filters only. Job titles, keywords, locations, seniority, and tenure persist; company-level filters have to be supplied again on each run.
- Sending requires a connected sender. Email needs a connected inbox, and LinkedIn messaging needs a connected LinkedIn account. The connector never sends LinkedIn connection invites on its own — those are only added automatically inside a campaign's outreach plan.
- Meetings are update-only. They can be modified through the connector but not created.
- Notes are author-scoped. Only the note's author can edit it.
- Discovery results expire. Saved searches are available for about 30 days.
Security and data handling
This section is for IT and security teams evaluating whether to approve the Strama connector. It summarizes the parts of our Security page that bear on the connector specifically; that page is the authoritative source and carries the full detail.
Authentication and authorization
- The server is a remote MCP server at
https://mcp.strama.ai/mcp, reached over HTTPS. It authenticates with OAuth 2.0 — no API keys or long-lived secrets are stored in the MCP client. - The connector operates as the authorizing user. It cannot read or modify anything that user could not already access in the Strama application.
- Access is scoped to the Strama teams the user belongs to. Team membership is verified server-side on every request, and permissions are role-based (Owner / Member). Every tool takes an optional team identifier and defaults to the user's default team; a team the user is not a member of is not reachable.
- OAuth tokens are stored encrypted. Access can be revoked from the MCP client's connector settings or from Strama, and revocation takes effect immediately.
What the tools can touch
- Read and write operations are in separate tools. No tool accepts a method parameter or proxies arbitrary requests.
- Read-only tools carry a read-only annotation and may run without per-call confirmation. Every write, send, and destructive tool prompts the user.
- The server calls Strama's own first-party API. It is not a general-purpose HTTP client and cannot be pointed at arbitrary endpoints.
- The tools do not read the client's memory, chat history, conversation summaries, or user files, and collect no conversation data beyond the arguments a tool needs to run.
- The data in scope is CRM data — contacts, companies, deals, campaigns, sequences, and the email and LinkedIn messages already in the team inbox — plus prospect data returned by discovery searches.
Outbound communication controls
- Only
send_email,send_linkedin_message, and starting a sequence viaset_sequence_stateresult in outbound messages. All three require user confirmation in the client. - Opt-out status is enforced server-side. A send to an opted-out contact is rejected regardless of how it was triggered.
- Per-sender daily send limits are enforced server-side and apply to connector-initiated sends the same as to sends from the application.
- Sends are attributed to a configured sender identity with a connected inbox or LinkedIn account. The connector cannot send from an identity the user has not connected.
- Every message sent through the connector lands in the team inbox and is visible in Strama's audit surfaces, the same as any other send.
Hosting, encryption, and retention
- Hosting. US-based infrastructure. The application runs on Vercel, the primary database is PlanetScale (MySQL on AWS), and file storage is Cloudflare R2 and Vercel Blob. Background processing runs through Google Cloud Tasks and AWS SQS.
- Encryption. TLS 1.2 or higher on all external traffic, with HTTPS enforced on every endpoint, and AES-256 at rest.
- Retention. Disconnecting an integration deletes its credentials immediately; records already synced remain until you delete them. On account closure, workspace data is removed within 30 days of request. Encrypted backups are retained on a rolling two-day basis.
- Access control. Production access is limited to Strama's two founders.
Sub-processors and compliance
- Sub-processors. Fourteen vendors process customer data, including Google Cloud for AI, Unipile for LinkedIn, Recall.ai for meeting bots, Deepgram for transcription, and Twilio for communications. We give 15 days' advance notice before adding a new one. The complete list is on the Security page.
- Compliance. Strama completed CASA Tier 2 in 2025, with a 2026 reassessment underway, and is a verified Google OAuth application under restricted-scope review. Strama does not currently hold SOC 2.
- GDPR and CCPA. Strama acts as a data processor and honors data-subject requests within 30 days.
- Incident response. Security incidents are acknowledged within 72 hours; vulnerability reports within three business days.
Full details, including the complete sub-processor list, are on our Security page. For how Strama collects and stores data generally, see the Privacy Policy. Security questions and vulnerability reports go to security@strama.ai.
Support
If something is not working or you want a workflow the connector does not cover yet, email support@strama.ai. Include the prompt you used and roughly when you ran it, and we can trace it.