Head to head · Agent harnesses · October 2026 research run

boxes.dev vs OpenMausBot

OpenMausBot scores 68.5 (B) on agent readiness against boxes.dev's 53.2 (D), and leads in 5 of 7 scored categories. boxes.dev leads on security & auth. Both do agent harnesses.

Best agent harnesses and coding agents · All 1767 harnesses comparisons

Which one, for what

boxes.dev D

Best for Developers who want Claude Code or Codex running on cloud copies of their dev environment, started from Slack, Linear, a schedule, an API call or another assistant.

Ahead on

  • Security & auth, 69 against 59

Also in its favour

  • A hosted endpoint, with nothing to install

Watch for

No status page, incident history or SLA found

OpenMausBot B

Best for A person or an outside agent who wants several CLI agents (Claude Code, Codex, Grok, ACP CLIs) organised as named bots with their own computers, routines and approvals, driven by MCP or a CLI.

Ahead on

  • Reliability, 60 against 32
  • Schema & documentation, 90 against 57
  • Agent ergonomics, 77 against 60
  • Payments & pricing, 75 against 50
  • Maintenance & community, 74 against 48

Also in its favour

  • Runs on your own machine
  • Free to start without a card
  • Open source

Watch for

New bots start at Full access, which approves provider permission requests without a card

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Schema & documentation OpenMausBot +33

boxes.dev 57

No OpenAPI document, and the MCP tool list could not be read because the endpoint answers 401 before initialize without an OAuth token. The automation API's request and response fields are documented in prose tables (5 of 25). llms.txt with every help page listed (10). The help pages say when to target a new devbox, an existing devbox or an existing thread, when to steer a busy agent and when to wake a devbox first (16). Field tables with enum values such as backend (codex or claude), delivery (steer or next) and four run statuses, and unknown fields are rejected. The MCP schemas themselves were not seen (9 of 15). curl examples and named errors such as 409 devbox_unavailable, 409 thread_unavailable and 503 devbox_wake_failed, with what each means (12). /v1 in the API path on the home page, a single dated product update of 27 October 2026 and no changelog (5 of 15). 5+10+16+9+12+5 = 57.

OpenMausBot 90

The MCP server's 20 tools each have a JSON Schema input with additionalProperties: false and required fields marked. No OpenAPI for the local HTTP API (25). Markdown docs in the repository, and the docs app's source serves llms.txt, llms-full.txt and Markdown routes. docs.openmausbot.com disallows every crawler, so the live llms.txt was not read (10). Descriptions state purpose and limits, such as that update_bot_profile cannot alter permissions, that send_bot_message may make the bot use external tools and that edit_bot_message is refused while a thread works, and the server instructions say to read before mutating. Few say when not to use a tool (15). Integer limits with minimum and maximum, an enum on scope, a pattern on send_id and oneOf shapes for responders (13). JSON configuration examples for each install path, the wait statuses and job states, CLI exit codes and a { ok, error, hint, exitCode } failure shape (12). MCP server version 1.1.0, release notes per version and a changelog page in the docs (15).

Reliability OpenMausBot +28

boxes.dev 32

Hosted reading, for the remote MCP server at api.boxes.dev/mcp and the automation run endpoints. No status page found. The site footer says "All systems normal" with no link, and the trust centre at trust.boxes.dev draws in a script our reader could not run (0). No readable incident history (5). No request rate limits found. Plans cap awake devboxes (3 on the trial, 5 to 25 on paid plans), and the API answers 409 active_fork_limit_reached at the cap (5 of 15). The automation API takes an Idempotency-Key and returns the original run on a retry with idempotentReplay: true, and the MCP connection retries a lost start or message with the same request key. No Retry-After or backoff guidance found (12 of 15). No SLA found (0). Neither the MCP connection nor the automation API is marked beta, and the product launched in June 2026 per its own blog (10). 0+5+5+12+0+10 = 32.

OpenMausBot 60

Read as local software. Official npm package openmausbot (Node.js 24 or newer) and signed, notarised macOS builds, a Windows installer and Ubuntu 24.04 packages, with supported platforms and a capability table in the README (20). 2,075 test files and a CI workflow on main. Of the last 23 finished CI runs on main on 30 October 2026, 10 passed, 3 failed and 10 were cancelled by newer pushes in a serialised queue (20). 365 open issues against about 3,940 issues and pull requests filed since 11 August, so most are closed, but the 25 newest open issues include several bug reports from 29 and 30 October with no reply yet (15). Release notes for every version in New, Easier and Fixed sections, but new capabilities ship in 0.1.x patch versions and no release marks breaking changes (5). Version 0.1.134, and the README calls the project early (0).

Maintenance & community OpenMausBot +26

boxes.dev 48

Closed service. CLI @boxes-dev/cli 1.0.1134 published on 17 November 2026 (30). Only one dated product update was found, on 27 October 2026, and earlier CLI versions' dates were not read, so three releases in 90 days are not established (0). A support address, an in-app chat and a blog, with no public tracker or changelog (8 of 25). A CLI but no SDK, and no entry in the official MCP registry was looked for (5 of 15). The CLI is published from GitHub Actions with npm trusted publishing and signatures (5 of 10). 30+0+8+5+5 = 48.

OpenMausBot 74

v0.1.137 on 1 November 2026, after v0.1.135 on 31 October and v0.1.136 on 1 November (30). 128 desktop versions from 11 August to 30 October 2026, and three more since (20). The maintainer has 4,221 commits and the contributor list returns at least 100 accounts. 587 issues and pull requests were open from about 3,940 filed on 30 October, when we could only sample the newest open issues, nearly all from the last two days and mostly unanswered so far (per the 30 October check; github.com refuses our requests on 2 November) (16). Not in the official MCP registry, which returns nothing for mausbot (0). CI on main runs on every push with most finished runs passing, npm provenance and an SBOM for the bundled Cua runtime (per the 30 October check) (8). 30+20+16+0+8 = 74.

Payments & pricing OpenMausBot +25

boxes.dev 50

No x402, MPP or L402 (0). Plan prices are public ($19, $99 and $249 a seat a month) and extra usage is $0.40 a box-hour, with a box-hour defined as one hour of a 4 vCPU, 8 GiB devbox (30). A free trial of 20 box-hours. The help page says some teams must add a payment method before starting, so a card-free start is not certain (20 of 30). A person signs in with Google or Apple and approves the OAuth grant or creates the automation secret in the desktop app (0). 0+30+20+0 = 50.

OpenMausBot 75

No payment protocol (0). MausBot Cloud, the hosted option, is priced by plan on a public page ($15, $49 and $99 a month, Team $169 and Business $599, or Business self-hosted $299 a month billed yearly, re-read on 2 November), with extra computer-hours at $0.05 each, so plan pricing (10). The app is free with no card or account (20). npx openmausbot installs and starts it with no signup (20). Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 50 is 75.

Agent ergonomics OpenMausBot +17

boxes.dev 60

The MCP tool count is unknown. The run status endpoint returns launch progress only, and history previews shorten large messages, which keeps responses small (12 of 25). Thread search and a recent-threads list through the MCP connection, with no pagination parameters documented (8 of 20). Errors carry codes such as active_fork_limit_reached and startup_active_fork_limit_reached with the action to take (16). Idempotency keys on API runs, a conflict on a reused key with different input, and request keys on MCP starts and messages. Tool annotations could not be read (16 of 20). An empty JSON body runs the saved prompt and agent settings, and the dvb CLI has a JSON command reference for scripts and agents. No SDK (8 of 15). 12+8+16+16+8 = 60.

OpenMausBot 77

20 tools, inside the 11 to 30 band, with annotations that separate reads from writes but no toolsets or read-only mode to load. Transcript reads are bounded and never return screenshot pixels (17). limit up to 200 messages, search capped at 100 hits, task_id and scope filters, and since_ms and after_version to keep watching a job (17). wait_for_conversation reports settled, needs-user, failed, stalled or timed-out with a redacted transcript tail, and the CLI's exit codes separate a pending approval from a timeout. We did not read every MCP error path (15). send_id makes a repeated send return the first receipt, wait is held by the server rather than polled, and every tool carries readOnlyHint, destructiveHint, idempotentHint and openWorldHint (20). Few required parameters and defaults for task, limit and timeout. The server answers only while the app or openmausbot serve runs, mutating tools need a paired token in packaged builds, and there is no client SDK beyond the CLI (8).

Security & auth boxes.dev +10

boxes.dev 69

OAuth for the MCP server, with protected-resource metadata naming mcp.boxes.dev and four scopes (threads:read, threads:write, events:subscribe, settings:write), a per-project grant and removal from settings. Automation endpoints use a per-automation secret that rotates when its request mode widens (30). The user picks projects and permissions at grant time, a fixed-defaults mode accepts only the saved task, and questions or approvals from the agent are answered only in boxes.dev, never through the MCP connection (15). Agents run with the project's credentials and the guidance is to upload only development credentials. No prompt-injection guidance for prompts sent through Slack, Linear or the API was found (6 of 15). Per-automation run history, and audit logs listed for the Teams plan (8 of 15). A valid security.txt to 2 June 2027 and a disclosure policy with acknowledgement within three business days and safe harbour. No bug bounty and no certification found (10 of 20). 30+15+6+8+10 = 69.

OpenMausBot 59

Harness reading of the framework checklist, as for the other harnesses. 30 for what leaves the machine by default, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. The harness binds to loopback and keys stay in ~/.openmausbot, but PostHog usage analytics are on until switched off, with a daily snapshot of what is set up, and the CLI guide says API keys are saved in plaintext with mode 0600 (18). Approval cards for provider permission requests, a narrow Always allow, isolated Local VMs and cloud computers, own-computer control off by default, and an MCP server that cannot approve, delete or touch credentials. New bots start at Full access, bot changes to skills, profiles and team setup apply without approval, and new bots keep every connected-app tool until grants are assigned (10). The docs say web page instructions never authorise credential use and that one-time codes go only to the site that sent them, and imported bots arrive with no grants, but bots read one-time codes from connected email without asking again (9). Per-thread NDJSON event logs, an Inspector, a decision log for connected-app grants and Undo lines for bot changes (12). SECURITY.md with an email, GitHub private vulnerability reporting enabled and scope notes that name one open risk. No published advisories, security.txt or bug bounty. npm releases use trusted publishing with provenance, and the Windows installer is unsigned (10).

Score by category

CategoryWeightboxes.devOpenMausBotEdge
Reliability16%3260OpenMausBot +28
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%5790OpenMausBot +33
Agent ergonomics13%6077OpenMausBot +17
Security & auth14%6959boxes.dev +10
Payments & pricing10%5075OpenMausBot +25
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%4874OpenMausBot +26
Transparency & trust7%6059boxes.dev +1
Negative events≤150-2
TotalD 53.2/100B 68.5/100

Facts side by side

Factboxes.devOpenMausBot
KindMCP serverAgent harness
VendorBoxes, Inc.Supamaus Software Private Limited
Hosted endpointhttps://api.boxes.dev/mcpno (local only)
TransportsStreamable HTTP, HTTPstdio
AuthOAuth or keyOAuth or key
PricingFreemiumFreemium
x402nono
LicenceProprietary. The service runs under the Common Paper Cloud Service Agreement 2.1, and the @boxes-dev/cli package is marked UNLICENSEDApache-2.0
Tools exposednone20
Read-only variant documentednono
llms.txtyesno
Last release2026-11-172026-11-01
Terms last updated2026-10-202026-10-10
Privacy policy last updated2026-10-212026-10-18
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingnot found in the textnot found in the text
Terms or service can change without noticenot found in the textnot found in the text
Arbitration or class-action waivernot found in the textnot found in the text
Popularitynone8.2k stars, 3.5k npm/wk

How the scores have moved

boxes.dev

No score change since it was listed.

OpenMausBot

Verdicts

boxes.dev

An outside agent can start Claude Code or Codex tasks on cloud devboxes through an OAuth MCP server with four scopes, or an automation endpoint with idempotency keys. No status page, SLA or tool schema could be read, and every route starts with a person setting up a project in the app.

OpenMausBot

A bounded 20-tool MCP server with typed schemas, annotations and idempotent sends, and a headless CLI with JSON output and exit codes, let another agent hand work to a bot team and wait on it. New bots start at Full access, so provider permission requests pass without a card, and usage analytics stay on until switched off.

Before you call either

boxes.dev

  1. Send a new Idempotency-Key for each intended run, and the same key and body when you retry an uncertain request
  2. Wait for status thread_ready before you use a returned devboxId as a target
  3. If you get 409 active_fork_limit_reached, free an awake devbox or wait. The plan's awake-devbox cap is full
  4. Read the thread in boxes.dev for the agent's answer. The status URL does not return it
  5. Ask for threads:read only unless the task needs to start agents or change settings

OpenMausBot

  1. Ask the owner for a pairing code from Settings > Phone, exchange it at POST /api/auth/pair, and set OPENMAUSBOT_TOKEN with an explicit OMB_PORT
  2. Pass a send_id with every send_bot_message and send_channel_message, so a retry returns the first receipt
  3. Use wait_for_conversation instead of polling, and on needs-user stop and tell the owner a card is waiting
  4. Check a bot's access level with the owner before sending work. New bots run at Full access unless defaults were changed
  5. Ask the owner to switch off Settings > General > Usage analytics if nothing beyond model calls should leave the machine

Questions

Which is better for AI agents, boxes.dev or OpenMausBot?

OpenMausBot scores 68.5 (B) on agent readiness against boxes.dev's 53.2 (D), and leads in 5 of 7 scored categories. boxes.dev leads on security & auth.

Can an agent call boxes.dev and OpenMausBot without installing anything?

boxes.dev has a hosted endpoint at https://api.boxes.dev/mcp. OpenMausBot runs on your own machine, with no hosted endpoint listed.

Are boxes.dev and OpenMausBot open source?

No open-source release is listed for boxes.dev. OpenMausBot is open source (Apache-2.0).

Other comparisons with boxes.dev or OpenMausBot

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.