boxd vs Modal (2026)

Agents on a machine act in outside systems through 128 integrations, including GitHub, Linear, Slack, Notion, Jira, Salesforce, HubSpot, Stripe and Zendesk, each with typed tools such as creating an issue or sending a message.

A person connects an account once through the provider's consent screen, personally or as a shared connection for the whole organization, and the platform then makes each call on the machine's behalf, so the credential never enters the machine.

The in machine run CLI and the @boxd/run SDK call these tools, and scripts reach any other API, database or cloud account with ordinary libraries.

Sourcedocs.boxd.sh/guides/integrations/connectionsread 2026-11-05

Code on Modal reaches other systems through what the customer writes. OIDC tokens authenticate functions to external clouds, cloud buckets mount as filesystems, and secrets hold credentials. Slack notifications report on apps (beta), and examples connect Discord, Google Sheets and Tailscale. Modal has no catalog of connectors, OAuth actions or tool gateway of its own, so the tools are the customer's code.

Sourcemodal.com/llms.txtread 2026-09-22

Background jobs give automations a managed lifecycle.

A script that registers a schedule or trigger runs as a job that survives reboots, waits for a missing connection instead of failing, and restarts with exponential backoff for up to five attempts before it is marked failed. boxd's fan out pattern has a coordinator create one machine per worker, run a coding agent headless in each and let a lead agent merge the reports, and a fork brings a machine's jobs along paused so nothing runs twice.

The sequencing, branching and hand offs in that pattern are code the customer writes against the SDK, and boxd has no workflow graph, step model or multi agent runtime of its own.

Sourcedocs.boxd.sh/guides/automations/jobsread 2026-11-05

Retries are first class on Modal. A Retries policy can sit on any function, alongside job queues, fan out with map and spawn, batch and dynamic batching, and gang scheduled clusters. A pipeline that mixes model calls with deterministic steps can be built on these pieces. Sequencing and branching live in the customer's Python. Modal has no multistep workflow primitive that records completed steps and resumes after failure.

Sourcemodal.com/docs/guide/retriesread 2026-09-22

Work reaches a machine on schedules, events and inbound requests. An automation registers durations or cron expressions with every() and subscribes to integration events with .on(), such as a new Linear issue or a pull request through the boxd GitHub App, and 36 integrations emit triggers.

The platform holds the next due time and wakes a hibernated machine shortly before it, holds events that arrive while a machine wakes for up to 15 minutes, and fires a missed cron slot once on wake. A hibernated machine also wakes on the next HTTPS request, so an agent served from it answers webhooks at its own address.

Sourcedocs.boxd.sh/guides/automations/scriptsread 2026-11-05

Scheduled Functions run on a cron expression or a fixed period. Web functions and endpoints boot a container from zero when a request arrives, and job queues hand work to functions as it lands.

Sourcemodal.com/docs/guide/cronread 2026-09-22

No retrieval index, vector store or knowledge base over the customer's content comes with the platform. Files on a machine's disk, standalone disks that outlive a machine and data sets saved into a snapshot are storage the agent's code reads, not an index a query returns. Tools such as Notion, Confluence or Google Drive are searched through their own integrations, and any grounding layer is something the customer installs on the machine.

Sourcedocs.boxd.sh/llms.txtread 2026-11-05

Retrieval systems the customer builds can run on Modal, and its examples embed documents with TEI and run a RAG chatbot over PDFs. Modal does not maintain a retrieval structure over the customer's content that an agent queries. Volumes, Dicts and mounted buckets are storage, and the product itself cannot index, refresh or permission sources.

Sourcemodal.com/llms.txtread 2026-09-22

Machines keep their working state between sessions. A machine on standby resumes in under a millisecond and a hibernated one wakes with its memory and processes intact, and coding agent sessions live on the persistent disk, so a job delegated in the morning can be resumed interactively later. Forks, checkpoints and snapshots copy or rewind that state.

Automations also get object(name), a named JSON store written on every assignment and restored on the next run, meant for cursors, dedup sets and counters. All of it is machine and script state the developer manages, with no memory layer of stated scope and lifetime that a buyer can review, edit or delete as memory.

Sourcedocs.boxd.sh/guides/automations/scriptsread 2026-11-05

State carries across runs. Sandbox snapshots save and restore a sandbox's filesystem and memory, memory snapshots speed cold starts, volumes persist files, and Dicts and Queues hold data shared across function calls. The agent's code reads that state to carry on. Restored process state belongs to the machine, and the customer can resume or delete it but cannot review or edit it. Modal has no memory layer for agents.

Sourcemodal.com/docs/guide/sandbox-snapshotsread 2026-09-22

A person can step into an agent's work while it runs. Every machine has a live desktop in the browser where its owner watches the agent's browser, takes over the keyboard to enter a password, solve a captcha or fix a form, and hands control back without stopping the agent. The preinstalled coding agents are told to share that desktop link whenever they need someone, so a login or approval happens in the browser the agent already opened. There is no approval gate, pause primitive or policy that holds an agent's action until a person reviews it.

Sourcedocs.boxd.sh/guides/desktopread 2026-11-05

Work running on Modal has no approval step, consent checkpoint or escalation to a person. Restricted Functions stop untrusted code from touching Modal resources, calling other functions or reaching Modal's APIs. With sandbox networking controls, they give isolation the developer configures, and no person reviews an agent's action.

Sourcemodal.com/docs/guide/restricted-accessread 2026-09-22

Organizations have three roles. Members use shared machines and create org billed ones, admins also invite and remove members, change roles, set defaults, create teams and revoke org API keys, and the single owner can demote admins. Admin invites need an interactive login, so an API key or a token minted inside a machine cannot send one.

People sign in with GitHub or Google, with SAML single sign on offered to enterprise customers on request. A DPA is available, and SOC 2 Type II and ISO 27001 are both in progress, not held. Each machine is a KVM microVM with its own kernel, and egress allowlists and host bound secrets are enforced on the host, outside the machine.

Sourcedocs.boxd.sh/guides/organizationsread 2026-11-05

A SOC 2 Type II audit report is available through Modal's Security Portal at trust.modal.com, and Modal supports HIPAA and has external penetration testing. Identity runs through SSO with Okta, Microsoft Entra or custom SAML, with SCIM provisioning (beta).

Access is managed with role based access control, user groups and service users, and an append only audit log records sensitive actions on Enterprise.

Secrets, Restricted Functions that cannot touch Modal resources, and sandbox networking controls round out the controls, and Sidecars (beta) keep credentials and harness logic in a separate container the agent's code cannot reach.

Sourcemodal.com/docs/guide/securityread 2026-10-23

Automations are visible and the agent's own steps are not. The console's Automations page lists every background job with its triggers in plain English, the integrations it uses, its state, last run and last error, plus a generated write up and flow diagram for each script, and run logs prints or follows a job's log from inside the machine. Regional uptime and incidents appear on a public status page. A coding agent's prompts, tool calls and outputs are not traced by the platform, and no audit trail of member actions is offered.

Sourcedocs.boxd.sh/guides/automations/jobsread 2026-11-05

The workload itself is recorded in detail. Modal keeps real time logs and metrics per function and container, GPU metrics, function stats, and an append only audit log of sensitive workspace actions on Enterprise. Audit logs, function logs and container metrics export to any OpenTelemetry provider or Datadog. Audit stays separate from runtime. Prompts, tool calls, retrieved knowledge and outputs can be inspected step by step only as far as the customer's own code logs them. There is no tracing of each call an agent makes.

Sourcemodal.com/docs/guide/otel-integrationread 2026-09-22

Each machine runs in Frankfurt or New York, picked per machine from the CLI, console or SDK or set as an organization default, and its disk, memory, snapshots, checkpoints and backups stay in that region on Latitude.sh bare metal. Scale customers can have boxd operate the cluster inside their own cloud account, and the same release self hosts on AWS, Google Cloud or any KVM host, free up to 16 worker vCPUs and licensed above that, with air gapped installs supported. boxd gives conflicting accounts of its US region, with New York live in some places and coming soon with a waitlist in others.

Sourcedocs.boxd.sh/guides/regionsread 2026-11-05

Workloads can be pinned to a region. Functions and Sandboxes take a region argument for the container region, dedicated Endpoints set both container and routing regions, and a data residency guide sets out per service what can be pinned and when customer data may leave the region. Region selection carries a price multiplier (1.15x broad, 1.75x narrow). Modal is a managed cloud with no self hosted option, and shared Endpoints do not take a region.

Sourcemodal.com/docs/guide/data-residencyread 2026-09-22

Machines arrive with Claude Code, Codex and OpenCode installed and configured through an instructions file that teaches each agent the platform, plus three preinstalled skills covering the in machine CLI, automations and the desktop, and a CLI skill for the laptop.

Guides walk through development workspaces, preview environments, CI runners, personal assistants, sandboxes, agentic SaaS, RL environments and agent swarms, and connectors exist for open source tools such as TanStack AI, Pydantic AI, Harbor and Scale AI AgentEnv. These are configured third party agents and recipes a developer builds from, not prebuilt agents of boxd's own.

Sourcedocs.boxd.sh/use-cases/coding-agentsread 2026-11-05

Ready made starting points come as a gallery of runnable examples and a Library of models deployable as Shared or Dedicated Endpoints. The agent examples include Cursor cloud agents, a background coding agent with OpenCode, a coding platform, a Claude Agent SDK Slack bot, a LangGraph agent in a GPU sandbox, a computer use agent over VNC and parallel evals. Developers build from these examples and models, and there is no browsable catalog of agents for a customer to adopt.

Sourcemodal.com/llms.txtread 2026-09-22

The customer decides which agents and models run. Claude Code and Codex use the customer's own subscription, signed in once and served to every private machine in the organization, or an API key stored as a secret, OpenCode signs in on each machine, and any other agent or model that runs on Linux can be installed. boxd adds no model of its own and does not route between providers, and it does not use customer data to train or improve any model.

Sourcedocs.boxd.sh/use-cases/coding-agentsread 2026-11-05

Model choice is the customer's throughout. Modal serves models from its Library on Shared Endpoints billed per token or Dedicated Endpoints with the customer's capacity and region. The customer can deploy any open model on GPUs with vLLM or SGLang, as its examples show for DeepSeek, Nemotron and others.

Sourcemodal.com/docs/guide/shared-endpointsread 2026-09-22

Outside code drives the platform through TypeScript and Python SDKs, a gRPC API whose full api.proto service definition is public for generating stubs in any language, and a CLI that returns JSON from its commands. Each region has its own API endpoint, and organization API keys keep working after the person who created them leaves. The SDKs cover machines, forks, snapshots, checkpoints, disks, egress, secrets, domains, organizations and teams, and a filesystem API reads and writes files inside a machine from code.

Sourcedocs.boxd.sh/reference/grpc-apiread 2026-11-05

Official SDKs in Python, JavaScript/TypeScript and Go (the latter two in beta) let outside code call Modal, alongside a full CLI that includes a skills command for coding agents. Web functions and endpoints expose the customer's code as HTTP APIs, and a gRPC API sits behind the open source client.

Sourcemodal.com/llms.txtread 2026-09-22

There are no test cases, scoring or quality gates for an agent's behavior. The platform is where evaluations run, not how they are scored. Forks give every reinforcement learning rollout or benchmark trial a byte identical starting state, Moyai ran 279,577 Harbor benchmark machines on it, and a reproducibility check compares two forks run with the same seed, but rewards, datasets and scoring belong to the customer's own framework.

Sourcedocs.boxd.sh/use-cases/reproducible-rl-environmentsread 2026-11-05

Evaluations run on Modal, but scoring them is left to other tools. Its example runs massively parallel evals with Harbor in Modal Sandboxes, and its reinforcement learning examples use verl and TRL, with the datasets, graders and scoring belonging to those frameworks. Modal has no fixtures, scoring, quality gates or comparison of agent runs of its own.

Sourcemodal.com/llms.txtread 2026-09-22

Agents drive a real browser on the machine's own desktop. Every machine ships a browser automation CLI built for agents, backed by Chromium with anti bot fingerprinting and a persistent profile so logins stick, and scripts get the same browser through Playwright. A machine can also drive a Chrome window on the user's laptop over the DevTools protocol, for sites that only accept the user's own network, and the preinstalled agents get a browser tool for it automatically. Any GUI app installed on the machine opens on the same desktop.

Sourcedocs.boxd.sh/guides/launch-a-browserread 2026-11-05

Computer use agents can run on Modal's machines. VM Sandboxes, now generally available, give an agent a full Linux computer that can run Docker stacks, graphical environments and mobile simulators. In one example, a Browser Use agent drives Chromium inside a Modal Sandbox, watched over VNC, with the model served from a Modal Endpoint. Control of the browser comes from Browser Use, a separate library. Modal has no API for screenshots, clicks or typing, so it supplies the environment but not the control.

Sourcemodal.com/blog/vm-sandboxes-agent-computersread 2026-10-23