Context EngineThe runtime for your agents
Run agents on the models you choose
Pick a model from every major maker or add your own. Each task runs in its own sandbox on your cluster, on a schedule if you want, until the work is done.
Maya Okafor09:20
Build the Q3 review deck from the updates doc and the valuation bridge. Same structure as Q2, the three flagged companies up front.
ModelsMachines
Search models
Context
CContext 1.5
Anthropic
Claude Opus 5Most capableClaude Sonnet 5BalancedClaude Haiku 4.5Balanced · 200K contextOpenAI
GPT-5.6 SolGPT-5.4 miniGemini
Gemini 3.1 Pro PreviewGemini 3 FlashFastKimi
Kimi K3Kimi K2.6Balanced · 262K contextDeepSeek
DDeepSeek V4 Pro
Qwen
QQwen 3.8 Max
EffortAuto
Ask anything
Maya Okafor09:20
Build the Q3 review deck from the updates doc and the valuation bridge. Same structure as Q2, the three flagged companies up front.
ModelsMachines
Search models
Context
CContext 1.5
Anthropic
Claude Opus 5Most capableClaude Sonnet 5BalancedClaude Haiku 4.5Balanced · 200K contextOpenAI
GPT-5.6 SolGPT-5.4 miniGemini
Gemini 3.1 Pro PreviewGemini 3 FlashFastEffortAuto
Ask anything
Choose the model for each task
Pick Context 1.5 or a model from Anthropic, OpenAI, Gemini, Kimi, DeepSeek or Qwen from the task's composer.
Maya Okafor09:20
Build the Q3 review deck from the updates doc and the valuation bridge. Same structure as Q2, the three flagged companies up front.
ModelsMachines
Search models
Context
CContext 1.5
Anthropic
Claude Opus 5Most capableClaude Sonnet 5BalancedClaude Haiku 4.5Balanced · 200K contextOpenAI
GPT-5.6 SolGPT-5.4 miniGemini
Gemini 3.1 Pro PreviewGemini 3 FlashFastKimi
Kimi K3Kimi K2.6Balanced · 262K contextDeepSeek
DDeepSeek V4 Pro
Qwen
QQwen 3.8 Max
EffortAuto
Ask anything
Give every task its own sandbox
Each task runs in a fresh sandbox on your cluster, as the person who asked, and the sandbox is destroyed when the run goes idle.
Activity3Draft slides from the Q2 template
Put recurring work on a schedule
Turn a task into an automation that runs on the schedule you set, with every run and its outcome kept.
Automations
New automationTasks that run on a schedule. Open one to see its runs, pause it, or change what it does.
ListCalendarRuns
Search automations
- Q3 valuation bridgeEvery weekday at 9:00 AM PST · Next Tomorrow 9:00 AMSucceeded 2h ago
- Q3 portfolio updatesEvery Monday and Wednesday at 8:30 AM PST · Next Thu 4:00 PMRunning
- LP letterEvery day at 6:00 PM PST · Next Today 6:00 PMSucceeded 22h ago
- Inbox triageEvery hour · Next Today 11:00 AMPaused
Use the models you trust
Choose from every major maker in one list, set your organization's defaults, and add your own endpoints.
Maya Okafor09:20
Build the Q3 review deck from the updates doc and the valuation bridge. Same structure as Q2, the three flagged companies up front.
ModelsMachines
Search models
Context
CContext 1.5
Anthropic
Claude Opus 5Most capableClaude Sonnet 5BalancedClaude Haiku 4.5Balanced · 200K contextOpenAI
GPT-5.6 SolGPT-5.4 miniGemini
Gemini 3.1 Pro PreviewGemini 3 FlashFastKimi
Kimi K3Kimi K2.6Balanced · 262K contextDeepSeek
DDeepSeek V4 Pro
Qwen
QQwen 3.8 Max
EffortAuto
Ask anything
Maya Okafor09:20
Build the Q3 review deck from the updates doc and the valuation bridge. Same structure as Q2, the three flagged companies up front.
ModelsMachines
Search models
Context
CContext 1.5
Anthropic
Claude Opus 5Most capableClaude Sonnet 5BalancedClaude Haiku 4.5Balanced · 200K contextOpenAI
GPT-5.6 SolGPT-5.4 miniGemini
Gemini 3.1 Pro PreviewGemini 3 FlashFastEffortAuto
Ask anything
Run tasks until the work is done
Every task runs as a durable workflow, in its own sandbox or on a machine you pair, and the board shows where each one stands.
Kanban
MineTeamBacklog2
Idle1
Running1
Done2
Failed0
No failures.
Kanban
MineTeamBacklog2
Running1
Put recurring work on a schedule
An automation runs a task on the schedule you set and keeps every run, so recurring work happens without anyone starting it.
Automations
New automationTasks that run on a schedule. Open one to see its runs, pause it, or change what it does.
ListCalendarRuns
Search automations
- Q3 valuation bridgeEvery weekday at 9:00 AM PST · Next Tomorrow 9:00 AMSucceeded 2h ago
- Q3 portfolio updatesEvery Monday and Wednesday at 8:30 AM PST · Next Thu 4:00 PMRunning
- LP letterEvery day at 6:00 PM PST · Next Today 6:00 PMSucceeded 22h ago
- Inbox triageEvery hour · Next Today 11:00 AMPaused
Automations
New automationTasks that run on a schedule. Open one to see its runs, pause it, or change what it does.
ListCalendarRuns
- Q3 valuation bridgeEvery weekday at 9:00 AM PST · Next Tomorrow 9:00 AM
- Q3 portfolio updatesEvery Monday and Wednesday at 8:30 AM PST · Next Thu 4:00 PM
- LP letterEvery day at 6:00 PM PST · Next Today 6:00 PM
- Inbox triageEvery hour · Next Today 11:00 AM
Secure by default
Every task acts as the person who asked, inside a sandbox on your cluster, and every step is a row in an append-only log.
Audit log
Who granted, shared and revoked access across the organization, and every agent permission decision, with who answered.
Search people, files, permissions and events
Network All actors Last 30 days | Time (UTC) | Outcome | Event |
|---|---|---|
| Today | ||
| 09:24:48.117 | Blocked | Atlas connected to api.factset.com:443 · not in the allowlist, nothing left |
| 09:22:31.774 | Allowed | Atlas connected to pypi.org:443 · in the allowlist |
Showing 1 to 2 of 2 events
Audit log
Who granted, shared and revoked access across the organization, and every agent permission decision, with who answered.
Search people, files, permissions and events
Network All actors | Time (UTC) | Outcome | Event |
|---|---|---|
| Today | ||
| 09:24 | Blocked | Atlas connected to api.factset.com:443 · not in the allowlist, nothing left |
| 09:22 | Allowed | Atlas connected to pypi.org:443 · in the allowlist |
Showing 1 to 2 of 2 events
Follow runs from any device
Start a task and follow its run from the web, the phone apps, the desktop app or the terminal, on the same models and the same runtime.
Kanban
MineTeamBacklog2
Idle1
Running1
Done2
Failed0
No failures.
Kanban
MineTeamBacklog2
Running1
Work across the Context platform
Workspace is where your people and agents work, Engine runs it, Unify connects it to your tools, and Evals makes it better.
- WorkspaceChat, files and tasks where your people and agents work. Every task started there runs on Engine.
- UnifyYour team's apps, internal services and files, connected so a run can use them with each person's own accounts.
- EvalsEvery run Engine finishes can be scored against your criteria, and failures inform the skill's next version.
- Secure VMThe sandbox each task runs in, with its network allowlist, command approvals and audit log.
Home
Find files, apps, people, anything…⌘K
RecentPinnedTasks
- q3-portfolio-review.pptx1h ago
- q3-valuation-bridge.xlsx2h ago
- q3-portfolio-updates.docx3h ago
- lp-letter.pptx5h ago
Working on Q3 portfolio review deck
Home
Find files, apps, people, anything…⌘K
RecentPinnedTasks
- q3-portfolio-review.pptx1h ago
- q3-valuation-bridge.xlsx2h ago
- q3-portfolio-updates.docx3h ago
- lp-letter.pptx5h ago
Working on Q3 portfolio review deck
Deploy in your environment
Run Engine on your own Kubernetes cluster through KOTS, with models reached through the gateway you choose.
- Your EKS or AKS clusterSandboxes and runs are pods on your own Kubernetes cluster, installed through KOTS.
- Your model gatewayReach models through the Context Gateway, LiteLLM, Amazon Bedrock, Azure OpenAI or a PrivateLink endpoint.
- Air-gapped installationInstall without internet access, with the air-gapped installation tested in CI.
- Your identity providerSign people in with SAML or OIDC and provision them with SCIM.
Context runs in production at Qualcomm. Read the case study
- Amazon EKS
- Azure AKS
- Air-gapped
Context Engine is included in every Context workspace, including the Free plan.
- Free$0Includes Context Engine
- Plus$20Per person per month
- EnterpriseCustomEnterprise deployments run in your own cloud.Talk to us about a plan
Questions
Which models can I use?
Context 1.5, the managed model, and models from Anthropic, OpenAI, Gemini, Kimi, DeepSeek and Qwen. An operator can add a custom model with its provider, model id and credentials, and test it without a redeploy.
What if a model has no zero-data-retention route?
On a deployment that requires zero data retention, a model with no zero-data-retention provider is shown as unavailable. It is never sent through another route.
Where does a task run?
In its own sandbox on your cluster by default. You can also pair a Mac, Windows or Linux machine from the terminal and pick it as the task's runtime.
What keeps a long task running?
Each task run is a durable workflow on Temporal, so a failed step is retried instead of losing the run.
Can a task run on a schedule?
Yes. An automation runs a task on the schedule you set, in the managed sandbox or on a paired machine, and keeps every run with its outcome.
Can we see how much capacity is left?
Yes. Settings > Infrastructure shows the pods, nodes, CPU, memory and storage of your deployment, and estimates how many more sandboxes fit.
