Skip to main content

ConnectKit

Purposefully built for AI Agents. ConnectKit gives agents access to user SaaS accounts — Gmail, Calendar, Drive, GitHub, Slack, and 520+ more — through one YAML file per service. OAuth, API keys, token vault, and tool discovery handled automatically. Used in production by the Executive Assistant agent system.

Embedded. Local. Open source. No cloud APIs, no hosted auth services. Runs entirely on-device with an encrypted SQLite credential vault + local YAML specs. Ships as a single Python package with zero external infrastructure dependencies.

Connect AI agents to SaaS. One YAML file per service. OAuth, token vault, tool discovery — handled automatically.

from connectkit import ConnectKitBridge

bridge = ConnectKitBridge(user_id="alice")
await bridge.discover()

# Agent gets tools for all connected services
tools = bridge.get_tool_definitions()
# → [google-workspace__gmail_list, github__issue_list, salesforce__soql_query, ...]

# Show the connector catalog
catalog = bridge.list_available()
# → [{name: "google-workspace", connected: true, setup_guide_url: "..."}, ...]

Why ConnectKit?

Every AI agent that needs access to user SaaS accounts ends up wiring three things together: OAuth flows, credential storage, and CLI/MCP tool wrappers. You build /auth/google/login, /auth/google/callback, token refresh logic, subprocess calls for gws gmail list... then you do it again for GitHub, again for Outlook.

ConnectKit does all of that once, done right. One YAML file per service. No Python code per connector.

Feature Status
524 SaaS connectors (from Nango's 779-provider baseline) ✅
YAML-based connector spec (no Python code per service) ✅
Encrypted SQLite credential vault (Fernet) ✅
Universal OAuth 2.0 router (one endpoint, all services) ✅
CLI adapter backend (wraps any SaaS CLI) ✅
MCP adapter backend (connects to MCP servers) ✅
Dual CLI + MCP support per connector (53+ connectors) ✅
Connector catalog (list available services + status) ✅
Agent meta-tools (list, connect, disconnect, health) ✅
Per-user credential isolation ✅
Self-hosted support (base_url for self-hosted instances) ✅
No external API dependencies (works offline) ✅

Installation

pip install connectkit

Core Concepts

One YAML file per service

# connectors/google-workspace.yaml
name: google-workspace
display: "Google Workspace"
setup_guide_url: "https://developers.google.com/workspace/guides/create-credentials"
auth:
  type: oauth2
  oauth2:
    authorize_url: https://accounts.google.com/o/oauth2/v2/auth
    token_url: https://oauth2.googleapis.com/token
    scopes:
      - https://www.googleapis.com/auth/gmail.readonly
      - https://www.googleapis.com/auth/calendar.readonly
  required_fields:
    - name: client_id
      label: "Client ID"
      input_type: text
    - name: client_secret
      label: "Client Secret"
      input_type: password
tool_sources:
  - type: cli
    command: gws
    install: npm install -g @googleworkspace/cli
  - type: mcp
    server_name: google-workspace
    command: gws mcp

Auth types

Type Use for Example
oauth2 OAuth 2.0 (Google, Microsoft, GitHub, Slack, 234+ services) Browser authorization → tokens vaulted
api_key API keys (Firecrawl, Stripe, Twilio, 288+ services) Paste key into form → vaulted
none No auth needed (local tools) Auto-connected

Tool descriptions (JSON Schema)

Each tool can define its parameters as a JSON Schema:

tool_descriptions:
  - name: google_workspace__gmail_messages_list
    description: "List recent emails from the user's Gmail inbox."
    parameters:
      type: object
      properties:
        max_results:
          type: integer
          description: "Maximum number of emails to return (default: 10)"
          default: 10
        query:
          type: string
          description: "Gmail search query (e.g. 'from:example@gmail.com', 'is:unread')"
      required: []

Legacy parameter_descriptions (flat string map) is still supported for backward compatibility, but parameters (full JSON Schema) is preferred — it gives the LLM better type hints, defaults, and enum constraints.

Backend Count Best for
CLI 63 SaaS with official CLIs (gws, gh, glab, m365, stripe, sf, vercel, etc.)
MCP 514 SaaS with MCP servers (first-party or community)
Dual (CLI + MCP) 53 Best available backend picked at runtime

CredentialVault

All tokens are stored in an encrypted SQLite database (Fernet encryption). One vault per user. Master key from CONNECTKIT_VAULT_KEY env var.

from connectkit import CredentialVault

vault = CredentialVault("./data/users/alice")
vault.store_token("google-workspace", "oauth2", {
    "access_token": "ya29...", "refresh_token": "1//..."
})
token = vault.get_token("google-workspace")

OAuth flow

# 1. User fills in client_id + client_secret → stored in vault via POST /connectors/connect
# 2. Flutter renders "Connect" button → opens:
#    GET /auth/login?service=google-workspace&user_id=alice
# 3. Browser redirects to Google OAuth → user authorizes
# 4. Google redirects to:
#    GET /auth/callback?code=...&state=...
# 5. Gateway exchanges code for tokens → stores in vault
# 6. Google Workspace: ✅ Connected

OAuth states are Fernet-encrypted and self-contained — any vault with the same key can validate them (10-minute TTL).

Self-Hosted Support

Services like Firecrawl, Sentry, GitLab offer self-hosted deployments. ConnectKit supports them with an optional base_url field:

required_fields:
  - name: base_url
    label: "Self-Hosted URL"
    placeholder: "https://firecrawl.example.com"
    input_type: url
    help_text: "Leave empty for cloud version"
    optional: true

Connector Catalog

Category Count Services
dev-tools 120+ GitHub, GitLab, Jira, Bitbucket, GitLab, Sentry, Datadog, PagerDuty, Snowflake, Vercel, Netlify, Postman, Figma
productivity 80+ Google Workspace, Microsoft 365, Notion, Airtable, Monday, ClickUp, Trello, Asana, Evernote
crm 50+ HubSpot, Salesforce, Pipedrive, Zoho CRM, Close, Copper
communication 40+ Slack, Discord, Zoom, Twilio, WhatsApp, Intercom, Zendesk
payment 30+ Stripe, Square, PayPal, Braintree, Adyen, Checkout.com
marketing 30+ Mailchimp, Klaviyo, SendGrid, Apollo, Brevo, Constant Contact
e-commerce 25+ Shopify, WooCommerce, Amazon Selling Partner, BigCommerce
analytics 20+ Mixpanel, Amplitude, PostHog, Tableau, Looker, Heap
hr 20+ BambooHR, Workday, Gusto, Personio, Greenhouse, Lever
storage 15+ Dropbox, Box, Google Drive, OneDrive, Egnyte
other 90+ Everything else

License

MIT — see LICENSE.

Author

Eddy Xu

Status

Alpha — actively developed. Core spec model, vault, OAuth router, and adapter backends are stable with full test coverage (152+ tests). 524 connectors shipped. Currently used in production in the Executive Assistant agent system.

Metadata

Release files for connectkit 0.1.5

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for connectkit 0.1.5
File Size Uploaded Provenance
connectkit-0.1.5.tar.gz 160.8 kB Without provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for connectkit 0.1.5
File Interpreter ABI Platform
connectkit-0.1.5-py3-none-any.whl Python 3 none any Details

Total release size: 489.6 kB

Release files / connectkit-0.1.5.tar.gz

Download URL connectkit-0.1.5.tar.gz
Size 160.8 kB
Tags Source
SHA-256 checksum
How to use checksums
3a3bef20789592856559f87733f1408d3a2e23353c99d274bebce4e44e032c24
BLAKE2b-256 checksum
How to use checksums
4bd4aa283fbd6700d4bd1b929186c514941488573eaa11c9b8c69adaa98230a4
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.9.15 {"installer":{"name":"uv","version":"0.9.15","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release files / connectkit-0.1.5-py3-none-any.whl

Download URL connectkit-0.1.5-py3-none-any.whl
Size 328.8 kB
Tags Python 3
SHA-256 checksum
How to use checksums
29cbe603048b4d49d8aeb8f4b450bb6ae048cab6b10409e690161a2db3a11b66
BLAKE2b-256 checksum
How to use checksums
83b1b86f6499787249a19c2e9de6933c3d654d21b9e9692c79e039b9afdf3b78
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.9.15 {"installer":{"name":"uv","version":"0.9.15","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Security for connectkit 0.1.5

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page