One complete API surface for cloud and local Computer Agents.
Every tenant-facing Computer Agents service is available through the versioned HTTP API and first-class TypeScript and Python clients. The platform UI, CLI, agents, external products, and local appliances all use the same service boundary.
Cloud and appliance configuration
Set the deployment once per client. Both SDKs accept an origin or a URL already ending in/v1 and apply the selected organization to normal requests, streams, uploads, and downloads.
https://api.computer-agents.comSend the API key as a Bearer token. API keys are account-bound and resolved against the selected deployment.
SDKs normalize the deployment URL. Direct HTTP integrations use the canonical /v1 prefix explicitly.
Pass organizationId / organization_id or send X-Computer-Agents-Organization on direct HTTP requests.
Execution and project work
Run agents on persistent computers and keep their work connected to projects, tickets, prompts, files, and thread history.
| Capability | TypeScript | Python | HTTP API |
|---|---|---|---|
Threads Conversation threads and message execution | client.threads | client.threads | 56 operations |
Agents Agent configuration and management | client.agents | client.agents | 12 operations |
Computers Persistent computer workspaces, runtime configuration, and container lifecycle | client.computers | client.computers | 54 operations |
Files File and directory operations on computer and resource workspaces | client.files | client.files | 15 operations |
Attachments Uploaded thread and workspace attachments stored in user computer workspaces | client.attachments | client.attachments | 3 operations |
Projects Project management and workspace grouping | client.projects | client.projects | 43 operations |
Tasks Task planning, board workflows, comments, sprints, releases, and task-to-thread execution | client.tasks | client.tasks | 24 operations |
Prompts Reusable prompt templates, versions, and prompt library management | client.prompts | client.prompts | 9 operations |
Knowledge and automation
Package reusable capabilities and operate durable workflows, queues, schedules, triggers, and multi-agent orchestration.
| Capability | TypeScript | Python | HTTP API |
|---|---|---|---|
Skills Custom user-defined skills with markdown instructions, files, and icons | client.skills | client.skills | 5 operations |
Knowledge Versioned Knowledge libraries and published document retrieval for people and agents | client.knowledge | client.knowledge | 18 operations |
Metronomes Agentic workflow automations composed from triggers, conditions, actions, services, loops, and notes | client.metronomes | client.metronomes | 43 operations |
Batches Durable, reorderable workload queue and manual shelf with capacity-aware release | client.batches | client.batches | 11 operations |
Schedules Time-based automations and recurring or one-time execution schedules | client.schedules | client.schedules | 8 operations |
Triggers Event-driven automations, webhook triggers, and trigger execution history | client.triggers | client.triggers | 9 operations |
Orchestrations Multi-agent workflows with reusable orchestration graphs and orchestration runs | client.orchestrations | client.orchestrations | 8 operations |
Quality, optimization, and release control
Test behavior, measure it, optimize exact saved candidates, enforce assurance policy, and promote only accepted versions.
| Capability | TypeScript | Python | HTTP API |
|---|---|---|---|
Guardrails Versioned hidden prompt adaptation sets that can be attached to agents | client.guardrails | client.guardrails | 15 operations |
Evaluations Versioned evaluation datasets and evaluation runs | client.evaluations | client.evaluations | 31 operations |
Tests Versioned test plans, deterministic test runs, results, evidence, artifacts, and worker leases | client.tests | client.tests | 16 operations |
Assurance Versioned release policies that bind Tests, Evaluations, and Agent Optimization evidence into auditable decisions | client.assurance | client.assurance | 14 operations |
Fine-Tuning Agent-led improvement jobs based on evaluation sets | client.fineTuning | client.fine_tuning | 11 operations |
Optimization Campaigns Durable bounded loops that propose, test, evaluate, assure, and promote exact saved candidates | client.optimizationCampaigns | client.optimization_campaigns | 12 operations |
Optimization Candidates Strict, content-addressed, saved-only Function and Metronome optimization candidates | client.optimizationCandidates | client.optimization_candidates | 8 operations |
Release Control Evidence-gated, content-addressed publication and deployment of accepted immutable candidates | client.releaseControl | client.release_control | 12 operations |
Products, data, security, and evidence
Deploy product services, operate structured data, scan repositories, review scientific evidence, and expose realtime voice agents.
| Capability | TypeScript | Python | HTTP API |
|---|---|---|---|
Resources Deployable and managed service surfaces exposed under /v1/servers | client.resources | client.resources | 60 operations |
Agent Runtimes Always-on managed agent APIs, runtime deployments, logs, and invocation | client.runtimes | client.runtimes | 15 operations |
Databases Managed structured data stores, collections, documents, and analytics | client.databases | client.databases | 15 operations |
Security GitHub repository security monitoring, exact-commit scan runs, findings, policy, and audit evidence | client.security | client.security | 40 operations |
Evidence Human review and controlled promotion of scientific evidence extracted by Equal Care agents | client.evidence | client.evidence | 5 operations |
Voice Agents Voice-agent configuration, sessions, calls, and realtime interaction controls | client.voiceAgents | client.voice_agents | 10 operations |
Organizations and platform operations
Manage tenancy, authorization, enterprise identity, account access, billing, communications, moderation, and deployment state.
| Capability | TypeScript | Python | HTTP API |
|---|---|---|---|
Organizations Organization tenancy, membership, and governance | client.organizations | client.organizations | 17 operations |
Teams Team memberships, invitations, and resource sharing inside a workspace | client.teams | client.teams | 18 operations |
Authorization Organization-scoped policy decisions, delegations, approvals, policy history, and decision audit | client.authorization | client.authorization | 10 operations |
Enterprise Identity Organization OIDC and Microsoft Entra identity connections, group mappings, and SCIM credentials | client.identityConnections | client.identity_connections | 11 operations |
API Keys Authenticated API key creation, listing, rotation, and revocation | client.apiKeys | client.api_keys | 5 operations |
Account Authenticated account identity and destructive account-data deletion | client.account | client.account | 6 operations |
Billing User budget, subscription, resource-billing preferences, and workspace inference settings | client.billing | client.billing | 35 operations |
Costs Cost analytics and thread-level usage breakdowns derived from ACP execution data | client.budget | client.budget | 4 operations |
Notifications Push notification token registration and device-token management | client.notifications | client.notifications | 6 operations |
Email Connected mailbox search, reading, and sending for the Email skill | client.email | client.email | 4 operations |
Reports Content reporting and moderation-report submission history | client.reports | client.reports | 2 operations |
Deployment and appliance integration
Target cloud or local appliances, bridge local workspaces, inspect deployment metadata, and monitor platform health.
| Capability | TypeScript | Python | HTTP API |
|---|---|---|---|
Local Bridge Local appliance discovery, workspace synchronization, commands, and bridge sessions | client.localBridge | client.local_bridge | 23 operations |
System Version and platform-level metadata endpoints | client.system | client.system | 8 operations |
Health Health and monitoring endpoints | client.health(), client.ready(), client.metrics() | client.health(), client.ready(), client.metrics() | 3 operations |
Intentional protocol boundaries
The few API-only surfaces are explicit. They are operator or provider protocols, not missing tenant SDK functionality.
Operator-only control-plane endpoints are intentionally not part of tenant SDKs.
SCIM is a standards-protocol integration surface consumed by identity providers, not an SDK resource manager.
Provider callbacks and internal dispatch routes (9)
/billing/admin/**Privileged billing-ledger and subscription repair operations are restricted to platform operators.
/billing/apple/verify-transactionThe signed StoreKit transaction receiver is used by the first-party platform client, not by general SDK consumers.
/internal/**Principal-session and execution-dispatch contracts are authenticated worker-to-control-plane protocols.
/metronomes/email/inboundInbound email receiver invoked by the connector provider.
/metronomes/telegram/inboundInbound Telegram webhook invoked by the connector provider.
/metronomes/github/inboundInbound GitHub webhook invoked by GitHub.
/webhooks/github/securitySigned Security GitHub App webhook receiver invoked by GitHub.
/webhooks/github/security/setupSecurity GitHub App installation callback invoked by GitHub in a browser redirect.
/threads/{threadId}/runs/{runId}/resume-dispatchInternal runtime handoff used by the durable execution dispatcher.
Parity is enforced, not assumed.
The OpenAPI document is the HTTP source of truth. A machine-readable capability manifest maps every public service tag to both SDKs, and the build fails if a tag is missing, unclassified, or stale. This page reads that same manifest, so the published matrix cannot silently drift from the SDK contract.