Niches · Agent harnesses

Run an always-on general agent reachable from chat with scheduled tasks

Two or three companies' docs show a full fit.

2companies, full fit
10companies, fit with a limit
101companies carry the tag
8share of the category's work
4first-hand accounts

What counts as done

The agent answers in the channel only to allowed senders, runs its scheduled tasks without supervision, and every tool call is logged where the owner can read it.

Measured, for our own tests: the scheduled tasks run unsupervised for a week. No docs we read state these figures, so they don't count against a listing here.

How agents ask for it: message my agent on Telegram and have it do things on my server, send me a morning digest, watch my notes and act on them, give the agent its own machine

Does the job 2

The docs show it meets the test above, and nothing we read says otherwise.

ListingWhy
Letta B
Letta
`letta server --backend local --channels telegram` serves a bot with `pairing`, `allowlist` or `open` DM policy, `letta cron` runs schedules, and the server example binds to loopback. source
Vibe-Trading C
HKUDS
`vibe-trading serve` runs Telegram and other channels with `allow_from` senders, pairing codes and an operator gate (star > allowlist > pairing > deny), scheduled runs deliver to channels, and an audit ledger records live actions. source

Does the job, with a limit 10

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

ListingWhy
Agent Zero D
Agent Zero, s.r.o.
Telegram, WhatsApp and email with allowed_users lists, a task scheduler, Tool Access policies and chat logs cover the job, but an empty allow list accepts everyone, the UI has no login until set, and Infection Check ships off. source
Clawboo B
Sanreds
A self-hosted server with routines, an access token and an audit log, reached through its own web app rather than Telegram or Slack; connectors reach outward through Composio, and no allowed-sender list exists. source
Cline C
Cline Bot Inc.
cline connect telegram, slack, discord, gchat, whatsapp or linear maps threads to sessions with access control, and cline schedule adds cron jobs, but the hub must stay on 127.0.0.1 or carry a room secret, and sessions run yolo. source
dsh-superagent (DeepSeek) E
DeepSeek (the LICENSE names the DeepSeek Harness Team; DeepSeek Harness's terms name Hangzhou DeepSeek Artificial Intelligence Co., Ltd.)
A main Session listens to GitHub and Feishu, the host enforces allowed lists and records Decisions, and a panel shows what needs a person, but Executors run with danger-full-access as the operator and there is no scheduler. source
goose BB
Agentic AI Foundation (Linux Foundation)
A gateway for chat platforms such as Telegram exists with GOOSE_GATEWAY_MAX_TURNS and a scheduler, but allowed-sender control and tool-call logging were not found in the pages read. source
OpenMausBot B
Supamaus Software Private Limited
Bots, routines, webhooks, approval cards and per-thread NDJSON logs cover the job, but the channel is MausBot's own app rather than Telegram, and access is by pairing token rather than a sender list. source
Paperclip C
Paperclip Labs, Inc.
Experimental Agent Chat with Slack, Discord, Telegram and AgentMail connectors, routines, approvals and an activity log exist, but the default `local_trusted` mode treats every request as the board admin and telemetry is on until opted out. source
Unstoppable E
Unstoppable Domains, Inc.
Scheduled workflows and Remote Control from a phone exist, but the agent answers in the app, not a chat channel, and the desktop must stay on. source
Verdent E
Verdent AI, Inc.
Telegram, Slack and Feishu bots pair to Manager and `/schedule` runs tasks, but only while the desktop app is active and with no sender allowlist beyond pairing. source
Warp (Warp Agent CLI and the Oz Agent API) C
Denver Technologies, Inc. (d/b/a Warp)
Factories take work from Slack, Linear, GitHub and schedules and reply in the thread, but on Warp's infrastructure rather than your machine. source

Couldn't check 2

The docs were closed to us, or silent. These carry no mark either way.

ListingWhy
Qoder CLI C
BRIGHT ZENITH PRIVATE LIMITED
docs.qoder.com answered 502 from the fetch proxy on three attempts and a DNS failure through WebFetch, so no page was read.
Qwen Code BB
Qwen team, Alibaba Group (no legal entity is named in the repository or the docs)
The docs list WeChat and Scheduled Tasks pages that were not read; the headless and approval pages describe no channel.

Doesn't do this job 42

Listings in the category that do something else, with the reason.

ListingWhy
Aider D
Aider AI LLC
No chat channel or scheduler. source
Amp C
Amp Frontier Corporation
No chat channel, scheduler or allowed-sender list is documented; Amp is a coding agent with a CLI, apps and orbs. source
Antigravity CLI C
Google LLC
No chat channel, scheduler or allowed-sender list; agy is a terminal and headless coding agent. source
Auggie CLI D
Augment Computing, Inc.
No chat channel or scheduler; automation means CI pipelines and GitHub Actions. source
AXLR BB
Underpass AI
No chat channel or scheduler is built in; a bot must call --exec or the service itself. source
Claude Code C
Anthropic, PBC
Channels are a research preview that -p ignores, scheduled work runs as cloud routines, desktop tasks or a session-bound /loop; no chat channel with an allowed-sender list or week-long headless daemon is documented. source
CLI Manager BB
solhun (Jeong Kyeonghun)
No chat channel or scheduler, and the app must be open. source
Conductor E
Conductor
No chat channel; the Mac app and the closing cloud are the only surfaces. source
Crush D
Charmbracelet, Inc.
No chat channel or scheduler. source
Cursor CLI E
Anysphere, Inc.
No chat channel or scheduler in the CLI. source
DeepSeek Harness (dsh) D
DeepSeek (the LICENSE names DeepSeek, and no legal entity is named in the repository or the docs)
No chat channel or scheduler in dsh itself. source
Devin C
Cognition AI, Inc.
A hosted coding agent for repositories, not a general assistant on the owner's machine; the Slack integration was not read. source
Droid CLI C
The San Francisco AI Factory, Inc.
No chat channel or scheduler; Missions orchestrate coding work. source
Pi B
Earendil Inc.
No chat channel or scheduler. source
Emdash C
General Action, Inc.
No chat channel; a desktop app. source
Gemini CLI B
Google LLC
No chat channel or scheduler. source
GitHub Copilot CLI C
GitHub, Inc.
No chat channel or scheduler for an always-on agent. source
Harness B
Autonomous, Inc.
No chat channel or scheduler. source
Harness² E
Google LLC
Not an assistant. source
HarnessRouter D
Lumentree Corporation
No chat channel or scheduler; a task API. source
Jules D
Google LLC
Not an assistant, and read-only since 11 November 2026. source
Junie CLI D
JetBrains s.r.o.
No chat channel or scheduler. source
Kilo Code CLI BB
Kilo Code Inc. (the terms are a contract with Anaconda, Inc. on behalf of itself and its affiliates, including Kilo Code Inc.)
No chat channel or scheduler. source
Kimi Code CLI C
Moonshot AI (the global Kimi terms name NOVASCENT PRIVATE LIMITED, Singapore)
No chat channel or scheduler is documented; the REST and WebSocket server is a local session API, not a messaging gateway. source
Kiro CLI B
Amazon Web Services, Inc.
The CLI has no chat channel or scheduler; Kiro Crew and Bots are separate surfaces not covered by this listing. source
mini-swe-agent B
No legal entity named. The MIT licence names Kilian A. Lieret and Carlos E. Jimenez
No chat channel or scheduler; it is a one-tool coding agent. source
MiniMax Code B
Nanonoble Pte. Ltd.
The CLI has no chat channel or scheduler; messaging integrations belong to the MiniMax Agent app, not `mcode`. source
Mistral Vibe CLI B
Mistral AI (RCS Paris 952 418 325)
No chat channel or scheduler in the CLI docs. source
Muse Code C
Meta Platforms, Inc.
No chat channel or scheduler is documented. source
OpenAI Agents API BB
OpenAI OpCo, LLC
No chat channel or scheduler; it is a session API. source
OpenAI Codex BB
OpenAI OpCo, LLC
No chat channel or scheduler in the CLI; Codex Cloud tasks start from ChatGPT surfaces. source
OpenCode BB
Anomaly Innovations, Inc.
No chat channel or scheduler; `opencode serve` is a session API. source
OpenHands BB
All Hands AI
No Telegram or Slack channel serving is documented; Slack appears only as an MCP target for automations. source
Orca C
Lovecast Inc.
No chat channel is documented; automations schedule prompts on repositories. source
Parallax E
Conifer Solutions, Inc.
No chat channel or scheduled tasks. source
Paseo B
Mohamed Boudra
Mobile and web clients talk to your daemon; there is no Telegram or Slack channel. source
Prime Agent C
Prime Intellect, Inc.
Slack is an MCP connector for tools, not a channel the agent answers in. source
Mirror CLI C
Reflection AI
No chat channel or scheduler. source
Rovo Dev CLI E
Atlassian Pty Ltd
No chat channel or scheduler in the CLI. source
Sculptor C
Imbue, Inc.
No chat channel or schedules. source
T3 Code C
T3 Tools Inc.
No chat channel or schedules; phone access is to your own server. source
Traycer C
Traycer AI
No chat channel or schedules; the mobile app follows your own tasks. source

What stops agents finishing

  • gateway listens on all interfaces with no auth by default (acc-30, acc-31)
  • cron jobs stop for two weeks unnoticed (acc-30)
  • a vague instruction deletes host files (acc-30)
  • the agent edits its own config, claims success and breaks startup (acc-29)
  • marketplace skills carrying malware (acc-30)

Around it

Before: messaging.receive. After: messaging.send, home-automation.control.

Where this comes from

Weight 8 of 100 in Agent harnesses: 4 of 33 accounts (acc-11, acc-29, acc-30, acc-31) all on OpenClaw, which is not a listing; Agent Zero and Letta are the listed equivalents

The same as data: /api/v1/use-cases/agent-harnesses.always-on-assistant.json, or the MCP tool get_use_case with id agent-harnesses.always-on-assistant. Other jobs in this category: /api/v1/niches?category=agent-harnesses.

Draft, read 2026-11-12 from public docs and first-hand accounts, not yet reviewed by a person. A mark says what the docs show, not what a run measured.

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.