01
Create your first agent
Start with instructions, connect a model, and give it the tools it needs.
OneAtlas
Docs
OneAtlas Platform
Build AI Products with OneAtlas
Ship production AI with simple primitives and enterprise-grade control.
Create Agent
Define behavior, tools, guardrails, and memory.
Build Workflow
Connect steps, approvals, and human handoffs.
Deploy App
Ship reliable AI products to production.
Monitor & Iterate
Trace runs, tune prompts, and improve outcomes.
Start
Get Started In Minutes
Move from idea to production workflow .
02
Build a workflow
Compose agents, actions, data, and review gates into one reliable flow.
03
Deploy to production
Publish to OneAtlas Cloud with observability and secure runtime controls.
What is OneAtlas
OneAtlas unifies agents, workflows, knowledge, memory and deployments into a single AI platform.
Agents
Autonomous task execution with typed tools.
Workflows
Durable orchestration for multi-step AI systems.
Knowledge
Ground answers in documents, apps, and APIs.
Memory
Persist context safely across users and sessions.
Deployments
Release, version, rollback, and scale apps.
Integrations
Connect your product stack with minimal friction.
Observability
Trace runs, evaluate quality, and monitor cost.
Security
Enterprise controls for access, audit, and data.
GET STARTED
How It Works
Every request follows a structured path from user input to production execution.
01
Receive Request
User submits a task or prompt.
02
Route Workflow
Workflow decides the execution path.
03
Call Tools
Agents access APIs, databases, and services.
04
Store Context
Memory and knowledge are updated.
05
Deliver Result
Response is returned and monitored.
BUILD
Agents
Build intelligent agents that reason, use tools, access knowledge, and complete tasks autonomously.
Reasoning
Analyze goals and determine the best execution path.
Tool Access
Connect APIs, databases, and external services.
Memory
Retain context across conversations and sessions.
Planning
Break complex tasks into manageable steps.
Human Handoff
Escalate workflows when review is required.
Safety Controls
Apply permissions, limits, and guardrails.
Agent Execution Flow
User Goal
→Agent Reasoning
→Tool Calls
→Memory Update
→Final Response
BUILD
Workflows
Orchestrate agents, tools, approvals, and business logic through structured execution flows.
Triggers
Start workflows from events, schedules, or user actions.
Conditional Logic
Route execution based on rules and outcomes.
Approvals
Insert human review checkpoints when needed.
Tool Execution
Invoke APIs, databases, and external systems.
State Management
Track progress and maintain workflow context.
Automation
Execute end-to-end processes without manual intervention.
Example Workflow
Lead Created
→Agent Review
→Score Lead
→Human Approval
→Assign Owner
BUILD
Tools
Extend agent capabilities with APIs, databases, webhooks, and custom functions.
Connected Tools
Available inside the runtime environment
Web Search
PostgreSQL
Browser
Custom Functions
Webhooks
Document Parser
Example
Agent uses multiple tools
User Query→Search API→Database→Agent Response
BUILD
Knowledge
Knowledge allows agents to access trusted information beyond their training data. By connecting documents, databases, and internal content, responses stay grounded, accurate, and relevant to your organization.
Retrieval Pipeline
Documents
→Indexing
→Vector Search
→Agent Context
→Response
Documents
PDFs, manuals, contracts, reports, and structured documentation.
Databases
Connect operational systems and structured business records.
Web Content
Ingest websites, help centers, and public knowledge resources.
Internal Sources
Sync team wikis, policies, playbooks, and private content.
Example
Grounded Customer Support Agent
Instead of relying only on model knowledge, the agent retrieves information from product documentation, support articles, and internal policies before generating a response.
User Question→Knowledge Search→Context Retrieval→Accurate Answer
BUILD
Memory
Memory enables agents to retain context across tasks and conversations. By combining short-term execution context with long-term stored knowledge, agents can provide more personalized, accurate, and consistent responses.
Short-Term Memory
Maintains active conversation context and execution state during a session.
- • Conversation history
- • Active workflow state
- • Tool execution context
- • Temporary variables
Long-Term Memory
Stores persistent information across sessions for personalized and adaptive experiences.
- • User preferences
- • Historical interactions
- • Saved knowledge
- • Organizational context
Memory Lifecycle
User Interaction
→Context Storage
→Memory Retrieval
→Personalized Response
BUILD
Evaluations
Measure how agents perform in production. Evaluations continuously monitor response quality, latency, cost, and reliability to ensure systems remain accurate and efficient as they scale.
Evaluation Dashboard
All Systems HealthyResponse Quality
94%
↑ 3% this week
Latency
1.2s
Within target
Tool Success
98.6%
Stable
Cost / 1K Runs
$4.21
Estimated runtime cost
DEPLOY
Deployments
Move agents and workflows from development to production in a few steps. OneAtlas handles packaging, versioning, rollout, and monitoring automatically.
Deployment Checklist
1
Configure agents and workflows2
Connect tools and knowledge sources3
Run evaluations and validation tests4
Choose deployment environment5
Deploy to production6
Monitor traffic and performanceDevelopment
Build & Test
Create workflows, connect tools, and validate behavior.
Staging
Verify Release
Run evaluations and simulate production traffic.
Production
Go Live
Release safely with monitoring and rollback support.
DEPLOY
Environments
Separate development, testing, and production workloads to safely release changes without impacting live users.
Release Promotion Flow
1
Development
BuildCreate agents, edit workflows, connect tools, and experiment freely. Changes are isolated and do not affect other users.
2
Staging
ValidateTest with production-like data, run evaluations, and verify tool integrations before releasing.
3
Production
LiveDeploy to live users with monitoring, scaling, and rollback support enabled automatically.
Recommended Workflow
Action
Development
Staging
Production
Build agents
✓
—
—
Run evaluations
✓
✓
—
User testing
—
✓
—
Serve customers
—
—
✓
Monitor traffic
—
—
✓
DEPLOY
Scaling
OneAtlas automatically distributes traffic across multiple workers to maintain performance as usage grows. Additional capacity is provisioned when demand increases and reduced when traffic drops.
What Happens During Traffic Spikes?
01
Traffic Increases
More users begin interacting with deployed agents.
02
New Workers Launch
Additional runtime instances are automatically provisioned.
03
Load Is Distributed
Requests are balanced across available workers.
04
Performance Remains Stable
Response times stay consistent under load.
DEPLOY
Observability
Monitor agents in production using logs, metrics, and execution traces. Observability helps identify failures, performance bottlenecks, and unexpected behavior before users are impacted.
[INFO] Agent initialized successfully
[TRACE] Knowledge search completed in 218ms
[WARN] External API latency increased
[ERROR] Payment service timeout detected
[INFO] Automatic retry completed successfully
What Can You Monitor?
- ✓ Agent execution history
- ✓ Tool calls and responses
- ✓ API latency and failures
- ✓ Memory retrieval performance
- ✓ Workflow execution paths
- ✓ Cost and token consumption
INTEGRATIONS
Databases
Connect your existing databases to agents and workflows. Once connected, data can be queried, updated, and used as context during execution.
Create Database Connection
1
Choose Database Type
Select the database you want to connect.
PostgreSQLMySQLMongoDBRedisVector DB
2
Provide Credentials
Configure host, port, database name, and authentication details.
3
Test Connection
Verify connectivity and permissions before activation.
4
Use in Agents & Workflows
Query data directly from agents or workflow steps.
INTEGRATIONS
APIs & Webhooks
APIs allow external applications to interact with OneAtlas programmatically, while webhooks enable OneAtlas to send real-time updates whenever important events occur.
API Request
POST /v1/agents/run
{
"agent": "support-agent",
"query": "Summarize today's tickets"
}
API Response
200 OK
{
"status": "success",
"summary": "...",
"tokens": 412
}
Event Flow
External App
→OneAtlas API
→Agent Runtime
→Webhook Event
Common Webhook Events
deployment.completedTriggered when a deployment succeeds.evaluation.finishedTriggered after an evaluation run completes.workflow.completedTriggered when a workflow reaches completion.agent.failedTriggered when an agent execution fails.INTEGRATIONS
Third-Party Tools
Connect the tools your team already uses. Integrations allow agents and workflows to interact with communication platforms, development tools, knowledge systems, storage providers, and business applications.
Integration Marketplace
120+ Available IntegrationsSlack
CommunicationSend alerts and interact with agents in channels.
GitHub
DevelopmentTrigger workflows from pull requests and commits.
Notion
KnowledgeUse workspace pages as agent knowledge sources.
Google Drive
StorageAccess documents, spreadsheets, and files.
Salesforce
CRMSync customer records and automate processes.
Jira
Project ManagementCreate and update issues directly from workflows.