Enterprise AI Agent Infrastructure

SDK + API + MCP

The governance layer
for your |

ThinkFleet is the infrastructure layer that makes AI agents trustworthy, governable, and enterprise-ready. Memory management, tool orchestration, credential governance, and guardrails — for Claude Code, Codex, Cursor, and any AI tool your team uses.

Request a Demo View SDK Docs

31 Patent-Pending Technologies400+ OAuth ProvidersFull Audit TrailPII Detection & RedactionSelf-Host or Cloud

Shield

AI proxy that governs every LLM request. PII detection, prompt injection defense, token budgets, and automatic memory injection.

OAuth Registry

400+ OAuth providers. Zero-knowledge credential proxy. Agents call APIs without ever seeing raw tokens.

SDK

14 resource modules. Memory, guardrails, orchestration, tools. Works with any AI tool via MCP, REST, or proxy.

thinkfleet-sdk.ts

import { ThinkFleet } from '@thinkfleet/sdk'

const tf = new ThinkFleet({
  apiKey: process.env.THINKFLEET_API_KEY,
  projectId: 'proj-abc',
})

// Guardrails — scan for PII before sending to LLM
const scan = await tf.guardrails.testScan({
  text: userMessage,
})

// Memory — search organizational context
const memories = await tf.memory.search(agentId, {
  query: 'deployment conventions',
})

// OAuth — agents never see raw credentials
await tf.oauth.proxy(
  'gmail/v1/messages/send', payload, connectionId,
)

// Shield — governance dashboard
const overview = await tf.shield.overview()

Ungoverned AI is your biggest risk

Every enterprise is adopting AI. The question is whether you govern it — or it governs you. ThinkFleet gives you the control layer your organization needs.

Shadow AI

Employees are already using ChatGPT, Copilot, and Claude with company data. Without governance, sensitive information leaks to third-party models with no audit trail and no controls.

Compliance Gaps

When AI makes a decision, who approved it? Without audit trails, PII detection, and prompt guardrails, every AI interaction is an untracked compliance risk.

Cost Sprawl

Teams spin up AI tools independently — each with its own billing, its own API keys, its own models. Token spend is invisible, unbudgeted, and growing fast.

Architecture

Governed AI from core to edge

Agents, crews, and flows at the center. Guardrails, audit, and access controls around them. Your infrastructure on the outer ring. Fully self-hostable.

ThinkFleetAI OS

Infrastructure

Interfaces

Core Pillars

PostgreSQL

pgvector

Redis

S3 Storage

Any LLM

Chat

Voice

API

Webhooks

MCP

Controlled Agents

Agent Crews

Automated Flows

Knowledge RAG

Guardrails

Audit & Observability

Access & Identity

Infrastructure

PostgreSQL, pgvector, Redis, S3, any LLM provider

Governed AI Core

Agents, crews, flows, knowledge, guardrails, audit, and access controls

Interfaces

Chat, Voice, REST API, Webhooks, MCP protocol

Your AI Workforce

Agents, Crews, and Flows — governed by design.

Three powerful capabilities for deploying AI across your organization — each with governance controls built in, not bolted on.

Controlled AI Agents

Deploy agents with guardrails built in

Deploy AI agents that operate within the boundaries you define. Every agent has token budgets, content moderation, PII detection, and prompt injection defense — enforced automatically on every interaction. Full audit trail for every action taken.

Capabilities

Multi-tool access across 660+ integrations
Streaming execution with real-time visibility
20+ pre-built expert personas
Knowledge base search with source citations

Governance Controls

Token budgets with daily per-project limits
PII detection and automatic redaction
Prompt injection defense (enabled by default)
Content moderation on input and output

Agent Crews

Patent Pending

Multi-agent teams with Kanban orchestration

Build teams of specialized AI agents that collaborate on complex tasks. A lead agent decomposes objectives, assigns tasks based on role and expertise, and tracks completion through a governed Kanban pipeline with cost tracking per task.

Capabilities

Automatic task decomposition with dependency resolution
Role-based agent assignment with round-robin fallback
Kanban pipeline with auto-advancement between columns
Per-task token accounting and cost estimation

Governance Controls

Exit rules validated before column transitions
Retry logic with configurable max attempts
Full execution state machine with cancellation support
Background sweep job recovers stale tasks automatically

Automated Flows

Visual workflow automation at enterprise scale

Build powerful automations with a visual drag-and-drop editor. Connect triggers and actions across 660+ services with branching logic, loops, and error handling. Every flow is versioned, auditable, and governed by the same controls as your agents.

Capabilities

Drag-and-drop flow builder with branching and loops
660+ pre-built integrations via secure OAuth
Schedule, webhook, and event-driven triggers
Real-time execution logs and debugging

Governance Controls

Version control with full rollback support
Audit logging for every execution
Multi-environment promotion (dev/staging/prod)
Flow-level execution history and replay

Your Governance Stack

Enterprise governance across every layer.

Guardrails, audit trails, access controls, and knowledge governance — applied consistently to every agent, crew, and flow in your organization.

Safety & Guardrails

PII detection, prompt defense, content moderation

Define what your AI can and cannot do with comprehensive safety controls. Content moderation, PII detection, prompt injection defense, and token limits ensure your AI workforce operates within the boundaries you set.

PII detection with BLOCK or REDACT actions
6-pattern prompt injection defense (enabled by default)
Input and output content moderation with severity levels
Per-project token budgets with daily enforcement
Tool restriction blocklists
Guardrail violation event logging

Observability & Audit

Real-time dashboards, metrics, and audit trails

Monitor your entire AI workforce from a single pane of glass. Real-time dashboards track conversations, token usage, errors, and costs. Audit logs capture every action — agent calls, flow executions, connection changes, and guardrail violations.

Real-time metrics: conversations, tokens, errors, cost
Audit events with IP address logging
HMAC-SHA256 event signing for tamper evidence
CSV export for compliance reporting
Configurable alerts with severity and acknowledgment
Automatic event retention and purge policies

Access & Identity

SSO, RBAC, MFA, and API key governance

Enterprise-grade access management with SSO (OIDC and SAML), role-based access control, multi-factor authentication, and scoped API keys. Everything you need to deploy AI safely across a large organization.

SSO via OIDC (Auth0, Keycloak, AWS Cognito)
SAML 2.0 SSO for enterprise identity providers
TOTP-based multi-factor authentication
Role-based access control (Admin, Editor, Viewer)
Scoped API keys with per-project permissions
IP allowlisting with CIDR support

Knowledge Governance

Controlled RAG with access policies

Upload documents, SOPs, and product manuals. ThinkFleet indexes everything with native pgvector embeddings — but access is governed. Per-team knowledge isolation ensures agents only access the data they are authorized to use.

Native RAG powered by pgvector embeddings
Multi-source ingestion: PDF, DOCX, web, Slack, databases
Source citations on every answer
Per-team knowledge isolation
Configurable chunk size and embedding models
Folder-based access organization

Agents + Crews + Flows + Governance = enterprise-ready AI

Patent Pending

Governed agent crews that ship real work.

Assemble teams of specialized AI agents. Import tasks from Jira, track cost per task, and watch your crew deliver through a governed Kanban pipeline — with exit rules, budget caps, and full audit trails.

Assemble Your Crew

Create specialized agent teams — each with a lead, roles, and tool access. Development, Marketing, Security — whatever your org needs.

app.thinkfleet.ai/crews

Crews page showing Development Team and Marketing Team with lead agents

Run the Pipeline

Agents work Jira-linked tasks through a Kanban board. Real-time cost tracking, per-task pricing, and automatic handoff between specialists.

app.thinkfleet.ai/crews/schedule-closings

Kanban board with Backlog, Requirements, Data Analyst, Security, and Development columns

Review AI Output

Every task produces structured deliverables — requirements docs, code reviews, security audits — with full conversation history.

app.thinkfleet.ai/crews/schedule-closings/task

Task detail showing AI-generated requirements document with recommendations

Learn More About Crews

Patent Pending Technology

Protected by 31 patent applications

ThinkFleet's AI orchestration, intelligent memory, and enterprise governance technologies are protected by patent-pending innovations — novel approaches you won't find anywhere else.

Agent Orchestration

5 patents

Hierarchical AI Agent Organization with Kanban Pipeline Execution

Patent Pending

Multi-agent orchestration through hierarchical org structures with Kanban pipelines, goal cascading, universal agent protocol, and dependency-aware task dispatching.

On-Demand Specialist Agent Creation via Personas

Patent Pending

AI dynamically creates specialized agents from persona templates, each with curated tool profiles and domain expertise.

Agent-to-Agent Communication via Crew State Machine

Patent Pending

Novel state-machine protocol enabling structured, deterministic communication between agents within a crew hierarchy.

Tool Discovery & Execution

5 patents

Connection-Driven Dynamic Tool Discovery

Patent Pending

Agents automatically discover available tools and capabilities based on active user connections — no manual configuration required.

Provider-Transparent Service Abstraction with JIT OAuth

Patent Pending

One unified interface for email, calendar, and CRM that transparently routes to Gmail, Outlook, or any provider with just-in-time authentication.

Shadow Flow Generation for Agent Tool Invocations

Patent Pending

Automatic creation of execution-backend flows for AI agent tool calls, enabling deterministic replay and audit of every agent action.

Memory & Knowledge

7 patents

Adaptive Impact Classification with Source-Trust Scoring

Patent Pending

Intelligent memory classification that scores extracted facts by business impact and source trust, automatically routing high-impact inferences through human confirmation.

Hierarchical Scoped Memory Retrieval and Inheritance

Patent Pending

Five-tier memory hierarchy — platform, project, agent, user, session — with automatic inheritance so new agents instantly know all institutional knowledge.

Token-Budgeted Memory Injection with Importance Weighting

Patent Pending

Dynamic allocation of limited context window space to persistent memory using confidence-based filtering, cross-scope deduplication, and importance-weighted compression.

Safety & Governance

11 patents

Transparent Proxy Memory Injection

Patent Pending

Zero-integration organizational memory injection into AI agent context windows via LLM API proxy. Works with Claude Code, Codex, Cursor without any client-side changes.

Zero-Knowledge Credential Routing via Authenticated Proxy

Patent Pending

AI agents invoke APIs through a proxy that auto-injects OAuth tokens. Agents never see raw credentials. 400+ provider registry with PKCE, automatic token refresh, and audit trail.

Heterogeneous AI Runtime Orchestration via Org Chart

Patent Pending

Enterprise org chart where each position can be backed by a different AI runtime (Claude Code, Codex, REST, desktop). Policy inheritance and budget enforcement across runtime boundaries.

Data Intelligence

3 patents

AI-Driven Cross-Database Query Planning

Patent Pending

Automatic relationship graph discovery from heterogeneous schema metadata, enabling natural-language queries across multiple connected databases.

Self-Healing UI Test Element Location

Patent Pending

Fixed-priority cascading selector hierarchy with AI vision terminal fallback that automatically repairs broken test selectors.

Multi-Channel Chatbot Auto-Configuration

Patent Pending

Automated webhook registration, heartbeat monitoring, and cross-channel context compaction for simultaneous deployment across messaging platforms.

Integrations

OS Pillars

Expert Personas

0.9%

Uptime

See it in action

One platform. Every capability.

From conversational AI to visual automation to multi-agent crews — explore the tools your AI workforce operates with.

app.thinkfleet.ai/tasks

ThinkFleet Task Hub — describe a task and the right AI expert handles it

Describe any task in natural language. ThinkFleet routes it to the right AI expert with your connected tools.

Why enterprises choose ThinkFleet