OurThinkTank/founders-os

šŸ¢ Workplace & Productivity
0 Views
0 Installs

šŸ“‡ šŸ  - Self-hosted business context for your AI client: CRM, check points, automation triggers, financial ledger, tasks, playbooks, news feeds, projects, and semantic memory in one MCP server. Ask one question and it reads your CRM and books together. Your data stays in your own database. MIT licensed.

Quick Install

One-Click IDE Configuration
claude_desktop_config.json
{
  "mcpServers": {
    "ourthinktank-founders-os": {
      "command": "npx",
      "args": [
        "-y",
        "ourthinktank-founders-os"
      ]
    }
  }
}
Or

Using an AI coding agent (Claude Code, Cursor, etc.)? Copy a ready-made prompt that tells it to fetch the setup instructions and install this server for you.

Documentation Overview

Founders OS

Open-source MCP server for startup and small business founders.

Founders OS gives you a complete business context - CRM, projects, tasks, finances, feeds, memory, playbooks - accessible from Claude, Cursor, or any MCP-compatible AI client. One connection, your entire business.

Built by OurThinkTank. Marketing site at foundersmcp.com.

New to AI tools, or handing this to someone who is? Start with Read This First - a plain-language intro to what FoundersOS is and isn't, what AI assistants can and can't do, and the habits that keep them honest.

Founders OS in action: typing 'start my day' produces a full morning briefing - tasks due today, work assigned to AI, and feed headlines.

What's Included

ModuleToolsDescription
CRM13Customers, contacts, interactions, pipeline dashboard
Tasks12Tasks with entity linking, AI assignment, dependencies, progress notes
Projects5First-class project records anchored on a project tag, with task rollups
Playbooks11Reusable orchestration templates that fan out to tasks and external MCP actions
Tags4Shared tag registry with soft validation and auto-registration
Financial14Double-entry ledger, P&L, multi-company, per-user access control
Feeds13RSS/Atom/JSON reader, briefings, bookmarks, pins
Memory5Semantic memory with personal + org scopes, pgvector, dedup, metadata filters
Surfaces6Cross-domain reads: session start, entity cards, weekly retro, stuck list, session checkpoints, project history
Members4Org membership directory, owner designation
Audit + Restore2Full audit log; soft-delete recovery
Diagnostic5Ping, version, usage guide, capability explorer, demos

94 tools total across 12 modules.

Quick Start

You need a Supabase project, an embedding API key (OpenAI by default), and an MCP-capable AI client.

1. Set up Supabase

Create a Supabase project, then in the SQL Editor run supabase/setup.sql. This single file sets up the full schema from scratch — extensions (vector, uuid-ossp, pg_trgm), all tables, indexes, RLS policies, functions, views, maintenance jobs, and the Data API grants required for compatibility with Supabase's removal of automatic default privileges for projects created on or after 2026-05-30. The wizard at foundersmcp.com/setup prints the same SQL with the embedding dimension already matched to your provider — prefer it if you are not using the default dimension.

2. Connect your AI client

The Founders OS MCP server runs through npx. Every client - Claude Desktop, Cowork, Cursor, Continue.dev, Zed, or any spec-compliant MCP client - uses the same configuration.

The quickest way is the wizard at foundersmcp.com/setup: enter your Supabase and embedding credentials and it generates a filled-in config for you to copy or download. Your credentials never leave the browser. You can also paste the block below by hand.

Drop this into your client's mcp.json (in Claude Desktop, this is the MCP servers section of your config):

{
  "mcpServers": {
    "founders-os": {
      "command": "npx",
      "args": ["-y", "@ourthinktank/founders-os@latest"],
      "env": {
        "SUPABASE_URL": "https://your-project.supabase.co",
        "SUPABASE_SECRET_KEY": "sb_secret_...",
        "FOUNDERS_OS_COMPANY_ID": "your-company",
        "FOUNDERS_OS_USER_ID": "your-name",
        "FOUNDERS_OS_TIMEZONE": "America/Los_Angeles",
        "EMBEDDING_PROVIDER": "openai",
        "EMBEDDING_MODEL": "text-embedding-3-small",
        "EMBEDDING_DIM": "1536",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

See Environment variables for the full list and provider options.

3. Try it

What can you do?                                  -> show_capabilities
Catch me up                                       -> get_session_start
Add Acme Corp as a new prospect                   -> add_customer
Log a call with Sarah at Acme - discussed pricing -> log_interaction
Create a task to send the proposal by Friday      -> create_task + link_task
What's stuck or overdue?                          -> get_stuck_list
Show me everything about Acme Corp                -> get_entity_card
Remember for the org: targeting SMB fintech in Q3 -> memory_store
Show me OTT's P&L for Q1                          -> get_pl_report
Run the customer-onboarding playbook for Acme     -> run_playbook
Give me my weekly retro for LinkedIn              -> get_weekly_retro
Let's checkpoint - wrap up this session           -> checkpoint
What's happened on founders-os lately?            -> get_project_history

CRM (13 tools)

Pipeline management for customer organizations and the contacts inside them. Customers are organizations; contacts are people - always separate records, so you can move a contact between customers without losing history.

Customers: add_customer, get_customer, update_customer, remove_customer, search_customers, list_customers

Contacts: add_contact, update_contact, remove_contact, search_contacts

Interactions: log_interaction, list_interactions

Dashboard: get_dashboard

Pipeline phases: prospect -> lead -> opportunity -> customer -> renewal (plus churned, inactive).


Tasks (12 tools)

Unified task management with org and personal scopes, entity linking, AI assignment, task dependencies, progress notes, and a task-to-memory bridge.

create_task, get_task, update_task, complete_task, remove_task, list_tasks, link_task, unlink_task, list_entity_tasks, add_task_note, assign_task, get_task_summary

Scopes: org (team-visible, default) and personal (private to creator).

AI assignment: Use @claude or @gpt as the assignee. get_task_summary surfaces a dedicated AI work queue. list_tasks(assigned_to='@claude') filters to AI-assigned work.

Dependencies: Set blocked_by_task_id on a task. Completing the blocker surfaces unblocked_tasks in the response.

Task-to-memory bridge: When completing a task, set store_as_memory=true to persist the completion note as an org-scoped memory entry.

Entity linking: Tasks can link to customers, contacts, interactions, transactions, projects, playbooks, memories, or any other entity type via the task_links junction table. Link at creation time or later with link_task.


Projects (5 tools)

Projects are first-class records anchored on a project tag (e.g. #acme-rebuild). get_project returns the project card with status, the linked tag, recent tasks grouped by status, and any customers tagged into it.

create_project, get_project, update_project, remove_project, list_projects

list_projects also flags any #-prefixed tags in the registry that don't yet have a project record, so the registry and the projects directory stay in sync.


Playbooks (11 tools)

Named, reusable orchestration templates. A playbook is defined once and run against a customer (or other subject) to spin up a complete project: it creates native Founders OS tasks AND, when connected MCP tools are present, fires external actions like creating a GitHub repo, posting to Slack, or scheduling a calendar event. If a connector is not available, the step gracefully falls back to a tagged [manual] task so the playbook still works.

create_playbook, get_playbook, update_playbook, remove_playbook, list_playbooks, add_playbook_step, update_playbook_step, remove_playbook_step, run_playbook, get_playbook_run, list_playbook_runs

Step types: native_task (Founders OS task) or external_action (MCP tool call). External steps carry a connector, an action, and a params object with placeholders.

Placeholders: {{customer.name}}, {{customer.slug}}, {{playbook.start_date}}, {{playbook.start_date+Nd}}, {{contact.primary.name}}, {{memory:key}} resolved at runtime.

Run log: get_playbook_run returns the full execution log; list_playbook_runs shows history per playbook.


Tags (4 tools)

Shared tag registry with soft validation. Tags are advisory: unrecognized tags warn but never block operations, and new tags auto-register on first use.

list_tags, create_tag, rename_tag, remove_tag

Conventions: #project-name for projects, @person for people, !state for meta-states (e.g. !needs-review). Simple category words like bug or release are fine unprefixed.

Validation checks: typo detection against existing tags, known-contact detection (nudges toward @), known-customer detection (nudges toward entity linking), and state-word detection (nudges toward !).


Financial (14 tools)

Simple double-entry ledger scoped by FOUNDERS_OS_COMPANY_ID, with per-user access control so company books can be opened to specific teammates only.

add_transaction, list_transactions, remove_transaction, add_category, list_categories, remove_category, add_account, list_accounts, remove_account, transfer_between_accounts, get_pl_report, get_financial_summary, get_financial_access, set_financial_access

Multi-company: Set a different FOUNDERS_OS_COMPANY_ID per instance to keep books separate.

Access control: set_financial_access grants or revokes a member's access to a company's financial tools. get_financial_access reports the current grants. By default the owner of a FOUNDERS_OS_COMPANY_ID has access; everyone else is locked out until explicitly granted.


Feeds (13 tools)

Built-in feed reader (RSS, Atom, JSON Feed) with a Postgres-backed store. Subscribe, brief, search, bookmark, pin.

Subscriptions: subscribe_feed, unsubscribe_feed, list_feeds, refresh_feeds, import_starter_feeds, pin_feed, unpin_feed

Items: get_feed_items, read_feed_item, get_feed_briefing

Bookmarks: bookmark_item, remove_bookmark, list_bookmarks

Categories: tech, startups, business, finance, product, design, engineering, ai, crypto, science, news, personal, other.


Memory (5 tools)

Semantic memory backed by pgvector with personal and org scopes, near-duplicate detection, metadata filters, and pagination.

memory_store, memory_recall, memory_update, memory_forget, memory_summarize_and_store

Scopes:

  • org - visible to all team members pointing at the same Supabase project
  • personal - visible only to the user whose FOUNDERS_OS_USER_ID matches

Filters on memory_recall: min_score, source_tool, created_after, created_before, offset, limit, project, scope.

Dedup: memory_store and memory_summarize_and_store check for existing memories with cosine similarity >= 0.92 and surface a conflict with options to force-store or skip.

Embedding providers (set via EMBEDDING_PROVIDER):

ProviderDefault modelDimsCredentials
openai (default)text-embedding-3-small1536OPENAI_API_KEY
bedrockamazon.nova-2-multimodal-embeddings-v1:01024AWS credential chain
ollamanomic-embed-text768OLLAMA_BASE_URL

Set EMBEDDING_DIM to match the model before running 002_memory_schema.sql. The dimension is permanent - changing providers later requires re-embedding the memory table.


Surfaces (6 tools)

Cross-domain read views that compose data from tasks, CRM, finance, and feeds into ready-to-render dashboards for AI agents.

ToolWhat it returns
get_session_startOrientation dashboard: task signals, AI queue, finance pulse, CRM activity, feed unread counts, suggested actions, first-run flag, and the four-tier rendering contract. Call at the start of every session.
get_entity_cardComplete picture of any entity (customer, contact, transaction, project) with open tasks, recent interactions, and linked records in one call.
get_weekly_retroCompleted-task retrospective grouped by tag with completion notes. Can format as a LinkedIn-ready draft.
get_stuck_listSurfaces stuck, stale, and overdue tasks that need triage, with days-stale counts and suggested actions.
checkpointEnd-of-session bookend to get_session_start. Returns the ordered wrap-up procedure (summarize, capture repo changes as commit links, store the record, propose follow-up tasks, write the handoff doc), the exact memory call to make, and the previous checkpoint so open items carry forward. Rides memory - no new entity, no migration.
get_project_historyChronological, newest-first timeline of a project's checkpoints; the "what happened, in order" companion to semantic memory recall. Pass kind: 'all' for every memory, not just checkpoints.

Members (4 tools)

Org membership directory. Maps FOUNDERS_OS_USER_ID slugs to display names, marks the owner of a company, and supports adding or removing members.

add_member, list_members, remove_member, set_member_owner

The owner of FOUNDERS_OS_COMPANY_ID is the default holder of financial access; others get access via set_financial_access.


Audit and Restore (2 tools)

get_audit_log returns the structured audit trail across all domains (creates, updates, deletes, restores, financial access changes, playbook runs).

restore_item reverses a soft delete on any soft-deleted record type, returning the record to its previous state. Use the audit log to find the original delete event and the entity ID to restore.


Diagnostic and Meta (5 tools)

ToolDescription
pingConnectivity test. Embeds an update notice if a newer package version is available.
get_versionRunning package version, rendering contract version, and the latest npm-published version.
get_usage_guideOn-demand reference covering modules, conventions, and common workflows.
show_capabilitiesFriendly overview with example prompts for each module.
list_demosLists or runs the bundled interactive walkthroughs (welcome tour, conflict resolution, run-my-week, etc.).

Rendering contract

Render-bearing tools include a render field with a four-tier ladder so the AI client picks the most visual output it supports: visual primitive tool (artifact/widget/canvas), inline rich output (HTML/SVG/JSX), markdown table, then prose. The contract ships in three channels in attention-strength order:

  1. Server instructions field at MCP registration - loaded at connect. All spec-compliant MCP clients.
  2. get_session_start.rendering_contract - full ladder text on session orientation. All clients.
  3. Per-response rendering_contract reminder - self-contained short form. Cold-start safety net.

The canonical source lives in packages/mcp-server/src/contract.ts, so every MCP client gets rich rendering through these channels with no plugin required. (A Claude plugin that mirrors the contract at system-prompt position, for stronger adherence in long sessions, is planned for a later release and is not part of this one.) Current contract_version is 4; a mismatch surfaces as contract_version_warning on get_session_start and get_version.


First-run onboarding

When the database is empty, tools attach onboarding hints to their responses. get_session_start detects a fresh install and suggests a guided walkthrough: add a first customer, create a first task, and optionally set up finance accounts. This keeps the experience conversational rather than dumping all 92 tools at once.

For a guided tour of a specific feature, ask the AI to "run the welcome demo" or "show me the conflict-resolution walkthrough" - list_demos returns the bundled interactive scripts.


Environment variables

# Supabase (required)
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SECRET_KEY=sb_secret_...

# Identity - set distinct values per teammate so personal memory scopes work
FOUNDERS_OS_USER_ID=your-name            # defaults to "default"
FOUNDERS_OS_COMPANY_ID=your-company      # defaults to "default"
FOUNDERS_OS_TIMEZONE=America/Los_Angeles # used by date-aware tools and YTD math

# Embedding provider for memory tools
EMBEDDING_PROVIDER=openai                # openai | bedrock | ollama
EMBEDDING_MODEL=text-embedding-3-small   # provider default used if omitted
EMBEDDING_DIM=1536                       # MUST match the vector() size in 002_memory_schema.sql

# OpenAI (required if EMBEDDING_PROVIDER=openai)
OPENAI_API_KEY=sk-...

# Bedrock (uses AWS credential chain - no key needed on AWS with IAM role)
# AWS_DEFAULT_REGION=us-east-1
# AWS_ACCESS_KEY_ID=...
# AWS_SECRET_ACCESS_KEY=...

# Ollama (required if EMBEDDING_PROVIDER=ollama)
# OLLAMA_BASE_URL=http://localhost:11434

Development

# Clone
git clone https://github.com/ourthinktank/founders-os.git
cd founders-os

# Install
npm install

# Build
npm run build

# Watch
npm run dev

Project structure

founders-os/
ā”œā”€ā”€ packages/
│   ā”œā”€ā”€ mcp-server/                  # @ourthinktank/founders-os npm package
│   │   ā”œā”€ā”€ src/
│   │   │   ā”œā”€ā”€ index.ts             # Entry point (stdio transport)
│   │   │   ā”œā”€ā”€ supabase.ts          # Database client
│   │   │   ā”œā”€ā”€ contract.ts          # Canonical rendering contract
│   │   │   └── tools/
│   │   │       ā”œā”€ā”€ crm/             # Customers, contacts, interactions, dashboard
│   │   │       ā”œā”€ā”€ tasks/           # Task management
│   │   │       ā”œā”€ā”€ projects/        # Project records
│   │   │       ā”œā”€ā”€ playbooks/       # Reusable orchestration templates
│   │   │       ā”œā”€ā”€ tags/            # Shared tag registry
│   │   │       ā”œā”€ā”€ financial/       # Ledger + access control
│   │   │       ā”œā”€ā”€ rss/             # Feed reader
│   │   │       ā”œā”€ā”€ memory/          # Semantic memory
│   │   │       ā”œā”€ā”€ surfaces/        # Cross-domain reads
│   │   │       ā”œā”€ā”€ members/         # Org directory
│   │   │       ā”œā”€ā”€ audit.ts         # Audit log
│   │   │       ā”œā”€ā”€ restore.ts       # Soft-delete recovery
│   │   │       ā”œā”€ā”€ diagnostic.ts    # Ping + version
│   │   │       ā”œā”€ā”€ meta.ts          # Usage guide + capabilities + demos
│   │   │       ā”œā”€ā”€ first-run.ts     # Empty-database hints
│   │   │       ā”œā”€ā”€ dates.ts         # Date/timezone helpers
│   │   │       └── permissions.ts   # Financial access checks
│   │   └── demos/                   # Interactive walkthrough scripts
ā”œā”€ā”€ integrations/
│   └── setup-page/                  # Config wizard hosted at foundersmcp.com/setup
ā”œā”€ā”€ supabase/
│   ā”œā”€ā”€ setup.sql                    # Complete schema for fresh installs (run once)
│   └── migrations/                  # Future schema changes (currently empty)
ā”œā”€ā”€ docs/                            # Specs and design docs
└── README.md

Local install for testing

Build, then point your MCP client at the local entry instead of npx:

{
  "mcpServers": {
    "founders-os": {
      "command": "node",
      "args": ["/absolute/path/to/founders-os/packages/mcp-server/dist/index.js"],
      "env": {
        "SUPABASE_URL": "https://your-project.supabase.co",
        "SUPABASE_SECRET_KEY": "sb_secret_...",
        "FOUNDERS_OS_USER_ID": "your-name",
        "FOUNDERS_OS_COMPANY_ID": "your-company",
        "EMBEDDING_PROVIDER": "openai",
        "EMBEDDING_DIM": "1536",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Contributing

Founders OS is open source under the MIT license. Outside contributions are not being accepted yet - that's coming soon. In the meantime, issue reports are welcome and very much encouraged: please file them on GitHub. See CONTRIBUTING.md for how to file a good report, and our Code of Conduct. Security issues should go through SECURITY.md, not public issues.

License

MIT - see LICENSE.

Related MCP Servers

6figr-com/jobgpt-mcp-server

šŸ“‡ ā˜ļø šŸ  šŸŽ 🪟 🐧 - MCP server for JobGPT — search jobs, auto-apply, generate tailored resumes, track applications, and find recruiters from any MCP client. 34 tools for job search, applications, resumes, and outreach.

šŸ¢ Workplace & Productivity0 views
Agentled/mcp-server

šŸ“‡ ā˜ļø - AI-native workflow orchestration with long-term memory, 100+ integrations, and unified credits. 32 MCP tools for building and running intelligent business workflows — lead enrichment, content publishing, company research, media production, and more. Knowledge Graph that learns across executions.

šŸ¢ Workplace & Productivity0 views
alex13slem/openproject-codex-plugin

šŸ“‡ ā˜ļø šŸ  šŸŽ 🪟 🐧 - Write-capable MCP server for OpenProject API v3 with Community Edition support. Search, create, update, assign, prioritize, and comment on work packages. Published as io.github.alex13slem/openproject in the official MCP Registry and installable with npx -y openproject-codex-plugin.

šŸ¢ Workplace & Productivity0 views
ap311036/ews-meeting-mcp

šŸ šŸ  šŸŽ 🪟 🐧 - Safely schedule Outlook meetings on on-prem Exchange EWS. Resolves attendees, discovers rooms, suggests slots, and requires preview-confirmed create/update/cancel writes with local credential handling and audit-friendly lifecycle records.

šŸ¢ Workplace & Productivity0 views

Engagement

Views
0
Installs
0
Upvotes
0

Views and upvotes are unique per visitor network (hashed IP). Installs count copy actions.

Status

Health: Not checked yet

We have not completed a health check for this listing yet.

No check timestamp yet.

Unclaimed listing (imported or pending owner verification). Claim it →
ā˜… Spotlight Slot

Feature Your MCP Server

Get maximum visibility for your server across our directory, search results, and detail pages.

Spotlight Your Server

Own this project?

This directory is pre-filled from public sources. Claim via GitHub README, site badge, or DNS TXT to get the verified badge and attach your website.

Claim this listing

Promote this listing

Optional paid placement. Free listings stay free forever.

Share & Embed

Add our SVG badge (dark/light directory styles) or embeddable widget to your site.