Give Claude a Team

Governed sub-agents. One shared workspace.

One MCP install, and Claude Code doesn't just call tools — it builds and deploys its own governed sub-agents. You and your agentic teammates share one workspace: the same connections, files, database, skills, and policies.

$npx @contextgate/mcp install

Works with Claude Code, Cursor, Windsurf, and any MCP client

Shared Workspace
ConnectionsFilesDatabaseSkillsPolicies
🧑‍💻
You
Claude Code
👤
Sarah
Teammate
🤖
Research Agent
Remote · Cloud
🤖
Sales Agent
Remote · Cloud
🤖
Ops Agent
Remote · Cloud
🧑‍💻
You
Claude Code
👤
Sarah
Teammate
↕
Shared Workspace
ConnectionsFilesDatabaseSkillsPolicies
↕
🤖
Research Agent
Remote · Cloud
🤖
Sales Agent
Remote · Cloud
🤖
Ops Agent
Remote · Cloud
🛠️

Claude spawns the agents

Ask in plain English and Claude Code calls the create-agent tool — building a governed sub-agent, its toolbox, and its policies in one step.

☁️

Hosted remotely, not on your laptop

Sub-agents run 24/7 on ContextGate Cloud — triggered by webhooks, schedules, or events. Close your terminal; they keep working.

🤝

Humans and agents, one workspace

People and agents are first-class members of the same workspace, sharing connections, files, the database, and skills — with a full audit trail.

Connections

One Place for Every Connection

Register MCP servers once. Credentials stored in your vault — never in your codebase or Claude config. Connect to 2,000+ apps out of the box.

Browse all 2,000+ connectors
Secret Backends
GCP Secret Manager, AWS Secrets Manager, Azure Key Vault
Health Monitoring
Automatic health checks and status tracking
Auto-Discovery
Sync tools and resources from MCP servers automatically
1000+ Apps
Composio integration for third-party OAuth connections
Shared Database

One Brain for All Your Agents

Every agent reads and writes to the same per-workspace DuckDB. One agent's output becomes another agent's input — automatically.

Sales Agent
Scrapes leads from LinkedIn
leads
Outreach Agent
Drafts personalized emails
Compliance Agent
Runs daily audit checks
audit_log
Reporting Agent
Generates weekly summary docs
Research Agent
Pulls competitor pricing
competitor_prices
Strategy Agent
Creates trend analysis report
Support Agent
Logs customer issues
tickets
Analytics Agent
Finds patterns, builds dashboards
Database
Query and manage your workspace data
v3 Undo Redo
SQL Query
DuckDB▶ Run
-- Sales Agent wrote 47 leads this week
SELECT company, contact_email, score
FROM leads
WHERE score > 80
ORDER BY score DESC LIMIT 5;
Results5 rows · 8ms
companycontact_emailscore
Acme Corpj.smith@acme.com97
TechStart Ltdcto@techstart.io94
Bright & Coops@bright.co.uk91
Nova Digitalhello@nova.dev88
Delta Servicesbd@delta.services85
Full SQL
SELECT, INSERT, CREATE TABLE, ALTER, DROP — powered by DuckDB
Version History
Automatic snapshots with time-travel restore
Workspace Isolated
Each workspace gets its own database — fully segregated
Central Skills Management

Teach Once. Every Agent Knows It.

Author a skill once — a playbook, an SOP, a reusable workflow — and every agent in the workspace can load it. Skills are version-controlled, governed, and managed in one place.

Skills
Shared playbooks available to every agent
📋
create-agent
Step-by-step playbook for building a new governed agent end to end
12 agents
🔎
diagnose-run
Audit a failed run from activity logs and request/response payloads
9 agents
🛡️
manage-policies
Author, update, and attach PII redaction and content-check policies
12 agents
🗂️
manage-workspace-data
Query, import, and snapshot the shared workspace database
7 agents
Update a skill once — every agent picks up the new version on its next run
Version-Controlled
Every skill edit is snapshotted — roll back any time
Workspace-Wide
Authored once, loadable by every human and agent in the workspace
Governed
Control which agents can author or edit skills, with a full audit trail

Claude Code is Powerful. Ungoverned, it's Dangerous.

No Visibility
Claude Code calls tools freely. You don't see what data flows where until it's too late.
Credential Sprawl
API keys scattered across .claude/ configs, env vars, and dotfiles. One leak away from disaster.
Zero Enforcement
No PII redaction, no compliance checks, no audit trail. Just raw tool calls hitting production APIs.
How it works

One Install. Full Control.

1

Install

One command adds ContextGate to your Claude Code MCP config. Works with any MCP-compatible client.

$ npx @contextgate/mcp install
2

Connect

Add your tools — Notion, GitHub, Slack, HubSpot, and 1000+ more. Credentials stored securely in your vault (GCP, AWS, Azure).

3

Govern

Define policies: PII redaction, blocked patterns, LLM-based checks. Every tool call input and output is enforced automatically.

4

Work

Claude Code uses 80+ tools normally. ContextGate intercepts, governs, and logs every call transparently. You stay in flow.

80+ Tools

Give Claude Minions

8 tool categories. 80+ managed tools. Every capability your agents need, governed by default.

37

Management

Create agents, manage toolboxes, configure policies, set up connections.

7

Triggers

Webhooks, cron schedules, event listeners, published chat interfaces.

5

Integrations

Connect OAuth apps (Slack, GitHub, Salesforce, Google Drive) and API key services.

6

File Management

Read, write, and search workspace documents — PDFs, DOCX, images, text files.

14

Browser Automation

Navigate, click, fill forms, screenshot — full Playwright-powered browsing.

7

Database

SQL queries, create tables, import data, snapshots, time-travel restore.

3

Run History

Agent activity analytics, full run details, token usage summaries.

2

Documents

Generate formatted Word docs, Markdown, CSV, JSON from agent data.

Policies

Compliance, Not Complexity

Pre-built templates for GDPR, HIPAA, and UK regulations. Or create custom rules with regex, PII entities, or LLM-based checks.

🔍

PII Redaction

NLP-powered detection blocks sensitive data before it reaches external tools.

EMAILPHONESSNCREDIT_CARDUK_NINOBANK_ACCOUNT
🚫

Block Patterns

Custom regex patterns and keyword lists to catch organisation-specific sensitive data.

RegexKeywordsCustom patterns
🤖

LLM Checker

Use an LLM to validate tool inputs/outputs against custom governance prompts.

Custom promptsCachedBidirectional
Policies
Manage PII redaction and governance policies
🔒 Policy Templates
3 available
Template
GDPR Strict
Redact direct identifiers...
gdprstrict
2 rules · EU, UK
Template
HIPAA Strict
Redact HIPAA direct...
healthcarestrict
2 rules · US
Template
UK Insurance
Redact UK insurance...
insuranceuk
2 rules · UK
Active Policies
Client Data Protection
Custom · 3 rules · Block mode
Active
pii_redaction — EMAIL, PHONE, UK_BANK_ACCOUNT, UK_NINO
block_patterns — /sort-code:\s*\d{2}-\d{2}-\d{2}/
llm_checker — "Reject if output contains financial advice"
Activity Logs

Every Decision, Logged

Every run from every human and agent lands in one searchable, filterable log — redactions, policy blocks, violations, and exceptions surfaced with full context for the audit trail.

5 of 12,847
Stop reason

stop_sequence

User

sarah@acme.com

Conversation

conv_7f3a2b91

Tools used

salesforce_create_account, hubspot_log_meeting

Toolboxes

Finance Ops Toolbox

Policies

Client Data Protection, GDPR

Redactions Applied
UK_SORT_CODE: 1IBAN_CODE: 1
Conversation
user

Add the new client profile to Salesforce with bank account 12-34-56 78901234 and IBAN GB29 NWBK 6016 1331 9268 19. Log the meeting in HubSpot.

assistant

Added the Acme Ltd account to Salesforce with bank details redacted, and logged the meeting in HubSpot.

Showing 1–5 of 12,847
Team & Cloud

One Workspace. Humans and Agents.

People and agents are members of the same workspace — sharing connections, files, the database, and skills. Deploy agents to the cloud and monitor everything from one dashboard.

🤝

Humans + Agents, One Workspace

Invite teammates and let Claude spin up agents — both are first-class members sharing the same connections, files, database, and skills.

☁️

Cloud Deployment

Deploy agents to ContextGate Cloud. Run 24/7 without your laptop open — triggered by webhooks, schedules, or events.

📊

Centralized Dashboard

See all agent runs, policy violations, and tool usage across people and agents in one place. Full audit trail.

🔐

Role-Based Access

Control who — and which agents — can create agents, modify policies, view audit logs, or manage connections.

Team
Manage workspace members
colleague@company.com
Invite
EmailRole
you@company.comOWNER
sarah@company.comMEMBER
contractor@ext.ioVIEWER
Governed Agents
3 agents deployed
DevOps Agent
82 tools · GDPR Strict
Active127 runs
Sales Agent
24 tools · Client Data
Active84 runs
Research Agent
12 tools · No policy
Active56 runs
All agents run 24/7 on ContextGate Cloud — no laptop required

Claude Code vs. Claude Code + ContextGate

Capability
Claude Code
+ ContextGate
Sub-Agents
None
Claude spawns governed cloud agents
Tools Available
Whatever you configure
80+ managed tools
Skills
None
Central, shared across every agent
PII in Tool Calls
Undetected
Blocked / Redacted
Credential Storage
Local dotfiles
Encrypted vault
Shared Database
None
Per-workspace DuckDB
Audit Trail
None
Full JSON-RPC log
Policy Enforcement
None
Block / Warn / Log
Browser Automation
None
Playwright-powered
File Management
Local only
Workspace-shared
Triggers
Manual only
Webhooks, cron, events, chat
Team Management
Single user
Shared workspaces + roles

Start Governing in 60 Seconds

One install. Full governance. From your terminal.

$npx @contextgate/mcp install
or get early access