MCP FOR AGENTS

Built from the Ground Up for Agents

357+ production tools. 5 agent profiles. Sentinel policy governance. Every capability exposed as an MCP tool call.

How It Works

Agents connect via MCP, authenticate with a profile-scoped key, then execute policy-checked tool calls with full audit logging.

1. Connect

Agent opens MCP session and discovers available tools based on its assigned profile and permissions.

2. Authenticate

Profile-scoped API key gates visible modules, operation limits, and Sentinel policy boundaries.

3. Execute

Calls pass Sentinel policy checks, optional confirmation gates, then full audit trail logging.

Agent → MCP Gateway → Sentinel Policy Layer → AgentFlow Engine

Tool Catalog

357+ tools grouped across 12 production modules.

Pipeline Orchestration

28 tools

PRD decomposition, pipeline templates, stage management, domain analysis

Session Management

32 tools

Session creation, pause, resume, handoff, progress tracking, context replay

Task & Sprint

45 tools

Task CRUD, sprint planning, velocity tracking, burndown, backlog grooming

Project Management

38 tools

Projects, epics, milestones, labels, cross-project linking, change requests

Agent Coordination

24 tools

Multi-agent execution, context assignment, handoff protocols, kill switches

Event Mesh

22 tools

Webhook endpoints, Kafka subscriptions, event filtering, outbound delivery

Sentinel Policy

28 tools

Policy CRUD, violation logging, allowlists, denylists, role-based assignment

Memory & Context

18 tools

Memory persistence, topic management, context assembly, knowledge retrieval

Audit & Compliance

16 tools

Audit queries, compliance evidence, actor attribution, exportable reports

Scheduling & Triggers

14 tools

Cron management, trigger creation, validation, history, deduplication

Local Sync

12 tools

Offline queue management, sync status, conflict resolution, cache control

File & Workspace

20 tools

File watching, intake processing, workspace management, output organisation

Agent Profiles

Five profile types map agent responsibilities to minimum required scope.

Profile	Primary Scopes	Best Fit
PIPELINE_EXECUTOR	Pipeline orchestration, session management, task execution	Assembly line agents executing multi-stage cognitive pipelines
PROJECT_WORKER	Task CRUD, sprint operations, project queries	Development agents managing tasks and sprint workflows
EVENT_RESPONDER	Webhook intake, event subscriptions, trigger management	Event-driven agents responding to external signals
AUDIT_READER	Audit queries, compliance evidence, read-only project data	Compliance agents generating reports and evidence packs
ADMIN_OPERATOR	Full access including policy management and system operations	Platform operations and controlled administrative actions

Profile

PIPELINE_EXECUTOR

Scopes

Pipeline orchestration, session management, task execution

Best Fit

Assembly line agents executing multi-stage cognitive pipelines

Profile

PROJECT_WORKER

Scopes

Task CRUD, sprint operations, project queries

Best Fit

Development agents managing tasks and sprint workflows

Profile

EVENT_RESPONDER

Scopes

Webhook intake, event subscriptions, trigger management

Best Fit

Event-driven agents responding to external signals

Profile

AUDIT_READER

Scopes

Audit queries, compliance evidence, read-only project data

Best Fit

Compliance agents generating reports and evidence packs

Profile

ADMIN_OPERATOR

Scopes

Full access including policy management and system operations

Best Fit

Platform operations and controlled administrative actions

Security Architecture

Every MCP call passes through Sentinel — the policy engine that governs what agents can access, modify, and execute.

Profile-scoped authentication

Sentinel policy enforcement

Per-agent rate limiting

Tool allowlist / blocklist

Dangerous operation confirmation gates

Scope and tenant isolation

Full MCP audit trail

Sensitive field masking

Agent Recipes

Outcome-focused playbooks across orchestration, operations, integration, and domain workflows.

PRD Assembly Line

Orchestration

Intent: Decompose a PRD into a full execution pipeline

Tools: pipeline_create, pipeline_decompose, task_create_batch

Profile: PIPELINE_EXECUTOR

Output: Domain spec, component outline, and ordered task list

Security Scan Pipeline

Domain

Intent: Run a multi-phase penetration testing engagement

Tools: pipeline_create, session_start, session_handoff

Profile: PIPELINE_EXECUTOR

Output: 7-phase security report with evidence chain

Insights Gap Analysis

Optimisation

Intent: Analyse session data and identify friction points

Tools: session_list, session_stats, memory_write

Profile: AUDIT_READER

Output: Quantified friction report with agent recommendations

Event-Triggered Build

Integration

Intent: Trigger pipeline from webhook or Kafka event

Tools: webhook_create, event_subscribe, pipeline_instantiate

Profile: EVENT_RESPONDER

Output: Automated pipeline triggered by external event

Sprint Status Reporter

Operations

Intent: Generate sprint status and velocity reports

Tools: sprint_get, task_list, sprint_velocity

Profile: PROJECT_WORKER

Output: Sprint health dashboard with burndown data

Compliance Evidence Pack

Compliance

Intent: Generate audit-ready evidence from execution history

Tools: audit_query, session_transcript, export_create

Profile: AUDIT_READER

Output: Timestamped evidence pack with actor attribution

Build Doctor Triage

Operations

Intent: Diagnose and fix broken builds iteratively

Tools: session_start, task_update, issue_create

Profile: PROJECT_WORKER

Output: Fixed build or escalation issue with diagnostics

Content Production

Domain

Intent: Transform brief into structured content output

Tools: pipeline_create, session_start, file_write

Profile: PIPELINE_EXECUTOR

Output: Outline, draft, and formatted content deliverable

Git Station Master

Operations

Intent: Handle mechanical git operations autonomously

Tools: session_start, task_complete, issue_create

Profile: PROJECT_WORKER

Output: Clean branches, resolved conflicts, or escalation

Intake Processor

Integration

Intent: Process file-drop change requests and issues

Tools: issue_create, task_create, file_move

Profile: EVENT_RESPONDER

Output: Created issue, draft task, and confirmation receipt

Integration Example

A minimal flow: connect via MCP, invoke a pipeline tool, then parse the structured result. Works with Claude Code, any MCP-compatible agent, or direct API calls.

const response = await fetch(
  "https://api.agentflow.dev/mcp",
  {
    method: "POST",
    headers: {
      "content-type": "application/json",
      "x-api-key": process.env.AGENTFLOW_API_KEY,
    },
    body: JSON.stringify({
      jsonrpc: "2.0",
      id: "req-1",
      method: "tools/call",
      params: {
        name: "agentflow_pipeline_run",
        arguments: {
          prd: "requirements.md",
          template: "api-assembly-line",
          auto_execute: true,
        },
      },
    }),
  }
);

const result = await response.json();

Ready to Connect Your AI Agents?

Send us your agent use case and we'll help you design the integration.

Request Early Access View All Features