agent-platform on tomrochette.com

AutoGPT

tom@tomrochette.com (Tom Rochette) — Sun, 27 Sep 2026 00:00:00 +0000

AutoGPT is the platform, frontend plus server, that Significant Gravitas rebuilt from the 2023 autonomous-agent phenomenon: today it is a hosted and self-hostable workflow runner where you describe an agent in plain English through AutoPilot or assemble it from blocks in a visual builder, then run it on demand, on a schedule, or from a trigger.

The star count says museum, the commit log says startup: AutoGPT is one of the few 2023 phenomena that survived by shipping a genuinely different product, and the accurate reading of its 187k stars is that they measure the meme, not the platform.

What it is
#

Four surfaces on one platform per the README: AutoPilot (chat a job into a working agent), Agents (mission control for every run, cost, and approval), Marketplace (ready-made community agents), and Build (the visual canvas). The stack is TypeScript and Python under the autogpt_platform directory, deployable on their cloud at platform.agpt.co or self-hosted from the repo. The old classic agent, Forge, and the AG Benchmark remain in the repo under MIT, but the platform is where development happens. It positions itself as “AI agents that finish the work”, with file-aware agents, scheduled and event-based triggers, and an integration block library.

Status
#

Active and shipping: the repo shows 187,582 stars and was pushed 2026-09-27, the day I checked (GitHub API). The platform releases on a cadence: beta v0.8.1 published 2026-09-24, v0.8.0 on 2026-09-19, v0.7.4 on 2026-09-04, v0.7.3 on 2026-08-28. The repo was created 2023-03-16 and the launch thread pulled 153 points with 174 comments in April 2023, at the peak of the autonomous-agent craze. License is dual: the platform folder is Polyform Shield (source-available, not OSI open source), everything outside is MIT. A Team plan is marked “coming soon”, which tells you the company is still mid-pivot from phenomenon to business.

Strengths
#

It is the rare zombie-brand resurrection that actually shipped: the same repo now runs a maintained platform with a real release cadence rather than a frozen meme.
The block-based builder plus chat-based creation covers both the control-hungry and the conversation-first users in one product.
Scheduling, triggers, and run visibility are first-class, which is what separates an agent platform from a demo.
Self-hosting is real and documented, so the Polyform Shield license does not lock out internal deployments.

Cautions
#

The name is now false advertising in both directions: people who remember AutoGPT 2023 expect an autonomous loop, and people who want an MIT platform will find Polyform Shield instead.
The 2023 heritage invites the old critique, and the “Auto-GPT Unmasked” thread’s production-pitfall skepticism still reads correctly today: hype compounds faster than reliability.
Platform folder under Polyform Shield means a competitor cannot build on it, which limits community forks of the part that matters.
A Team plan listed as “coming soon” with contact-sales pricing signals an enterprise motion still under construction.

Pricing
#

Hosted plans are credit-based: Pro at $42.50 per month billed annually (1x usage) and Max at $272.00 per month billed annually (8.5x usage), with yearly billing saving 15 percent versus monthly (as of 2026-09-27). A Team plan is contact-sales and marked coming soon; self-hosting the platform is free under the licenses above. An instant demo with no signup is offered.

Price history
#

Date	Plan	Change	Source
2026-09-27	Pro / Max	Baseline recorded: $42.50 / $272.00 per month billed annually	https://agpt.co/pricing

Compared to
#

LobeHub: the other hosted agent-operations platform; LobeHub grew out of a chat client toward scheduling, AutoGPT grew out of an autonomous loop toward structure, and they will meet in the middle.
Vibe Kanban: choose it for developer-centric parallel coding agents; choose AutoGPT for non-coding business workflows with schedules and triggers.
Omnara: the open control plane that supervises agents you already run, where AutoGPT wants to own the whole build-and-run loop.

Bottom line
#

Recommended for teams that want hosted, scheduled, observable agent workflows and accept a source-available license for the platform. Not for developers who need an OSI-licensed foundation to build a product on, and not for anyone expecting the 2023 autonomous experience. My disagreeable claim: the 187k stars are a liability now, because they attract nostalgia traffic the product cannot serve and bury the platform it became.

Changes
#

2026-09-27 - Created when the owner’s GitHub-stars scan surfaced it.

References
#

https://api.github.com/repos/Significant-Gravitas/AutoGPT - GitHub API (200): 187,582 stars, pushed 2026-09-27, Python, created 2023-03-16 (as of 2026-09-27)
https://raw.githubusercontent.com/Significant-Gravitas/AutoGPT/master/README.md - four surfaces (AutoPilot, Agents, Marketplace, Build), platform pitch (200)
https://raw.githubusercontent.com/Significant-Gravitas/AutoGPT/master/LICENSE - Polyform Shield for autogpt_platform, MIT for the rest (200)
https://github.com/Significant-Gravitas/AutoGPT/releases - releases (200): platform beta v0.8.1 published 2026-09-24
https://agpt.co/pricing - Pro $42.50 and Max $272.00 per month billed annually, Team coming soon (200, 2026-09-27)
https://docs.agpt.co - AutoGPT Platform documentation (200)
https://agpt.co/blog/introducing-the-autogpt-platform - the platform introduction post the LICENSE points to (200)
https://hn.algolia.com/api/v1/items/35413054 - launch thread (200): 153 points, 174 comments, 2023-04-02
https://hn.algolia.com/api/v1/items/35562821 - “Auto-GPT Unmasked: The Hype and Hard Truths of Its Production Pitfalls” (200): 90 points, 2023-04-13 (critical source)

LobeHub

tom@tomrochette.com (Tom Rochette) — Sun, 27 Sep 2026 00:00:00 +0000

LobeHub is the TypeScript platform (formerly LobeChat) that repositioned itself from open-source AI chat client into a “Chief Agent Operator” which, in its README’s words, organizes your agents into 7x24 operation by hiring, scheduling, and reporting on your entire AI team.

LobeHub’s 82k stars measure demand for a polished multi-provider chat client, not for agent operations, and the Chief Agent Operator story is a renaming of that demand rather than evidence for it.

What it is
#

A Next.js TypeScript application you can self-host (Docker, desktop apps) or use on LobeHub Cloud, with a plugin and skills ecosystem the README counts at 10,000+ tools and MCP-compatible plugins. The agent-operations layer is new positioning on top of the chat product: assistants you “hire” get scheduled to run recurring work, and a reporting view shows what your agent team did while you were offline. It speaks to many providers out of the box, and its pricing page shows GLM, GPT, Claude, and Kimi models metered through a credits system. Deployment spans cloud, self-hosted servers with private models, and desktop canary builds. The vendor is LobeHub LLC, which grew the product from the LobeChat project created in May 2023.

Status
#

Active and shipping fast: the repo shows 82,839 stars, about 15.9k forks, and a push on 2026-09-27, the day I checked (GitHub API). The release train is canary-grade desktop builds published several times a day; v2.2.19-canary.28 landed 2026-09-26 and five more canaries followed within hours. The rebrand from LobeChat is visible in the license text itself, which reads “From 1.0, LobeChat is licensed under the LobeHub Community License”. Community footprint outside its own channels is thin: the best HN thread about it (as LobeChat) reached 7 points in July 2024, and my HN search for the current name returned nothing above noise. A Product Hunt “top post” badge for the multi-agent-work launch and a Trendshift badge suggest its audience lives on Product Hunt and Chinese-language channels, not HN.

Strengths
#

The self-hosted, bring-your-own-model base is real and polished, with cloud, Docker, desktop, and private-model paths that few agent platforms match.
A 10,000+ skills and plugin ecosystem gives scheduled agents something to actually do on day one.
Multi-provider by design, so model churn lands on their pricing page instead of your migration backlog.
Release velocity is unusual; the canary cadence means fixes arrive in hours, not quarters.

Cautions
#

The license is not open source: the LobeHub Community License is Apache 2.0 plus conditions that allow commercial use only without modifying the source, and require a commercial license for derivative works.
The same license grants the producer the right to change terms and to use contributed code commercially, a one-way contributor bargain.
The agent-operations layer is young positioning; I found no independent evidence, benchmarks, or case studies that the “hire, schedule, report” loop works at production depth.
HN-scale scrutiny is nearly absent, so the safety question a commenter asked in 2024, how safe is this, still has mostly vendor-provided answers.

Pricing
#

Cloud tiers meter usage in credits (as of 2026-09-27): Free with 500,000 credits per month, Starter at $9.9 per month (5M credits), Premium at $19.9 (15M), and Ultimate at $39.9 (35M), with yearly billing saving 20 percent versus monthly rates of $12.9, $24.9, and $49.9. Self-hosting is free to run against your own provider keys, with private-model and custom integration options behind a contact link.

Price history
#

Date	Plan	Change	Source
2026-09-27	Starter / Premium / Ultimate	Baseline recorded: $9.9 / $19.9 / $39.9 per month billed yearly (monthly $12.9 / $24.9 / $49.9)	https://lobehub.com/pricing

Compared to
#

AutoGPT: the closest hosted rival; AutoGPT builds structure around workflows, LobeHub wraps operations around chat agents, and LobeHub’s license is less restrictive for commercial use.
Vibe Kanban: the developer-side scheduling board; choose it when the agents are coding agents in worktrees, LobeHub when they are general assistants.
Omnara: the open control plane for supervising agents you already run, closer to LobeHub’s reporting story but developer-first and Apache-style licensed.

Bottom line
#

Recommended for individuals and teams who want a beautiful self-hosted home for multi-provider assistants with light scheduled automation, and who accept a source-available license. Not for anyone who needs an OSI-licensed foundation, a commercial derivative product, or production-grade evidence behind the agent-operations claims. My disagreeable claim: the hire-and-schedule framing will not hold, because the hard problems in 24/7 agent operations are permissions, credentials, and verification, none of which a chat client’s architecture makes easy.

Changes
#

2026-09-27 - Created when the owner’s GitHub-stars scan surfaced it.

References
#

https://api.github.com/repos/lobehub/lobehub - GitHub API (200): 82,839 stars, pushed 2026-09-27, TypeScript, created 2023-05-21 (as of 2026-09-27)
https://raw.githubusercontent.com/lobehub/lobehub/canary/README.md - Chief Agent Operator pitch, 7x24 hire/schedule/report language, 10,000+ skills (200)
https://raw.githubusercontent.com/lobehub/lobehub/canary/LICENSE - LobeHub Community License text: Apache 2.0 plus commercial conditions (200) (critical source)
https://lobehub.com/pricing - Free/Starter/Premium/Ultimate tiers, credit system, self-host and private-model options (200, 2026-09-27)
https://github.com/lobehub/lobehub/releases - releases (200): Desktop Canary v2.2.19-canary.28, published 2026-09-26
https://lobehub.com/docs - documentation root (200)
https://hn.algolia.com/api/v1/items/41070091 - the main LobeChat HN thread (200): 7 points, 2024-07-25, includes the creator’s design-engineer answer and a safety question (critical source)
https://hn.algolia.com/api/v1/search?query=lobehub&tags=story - search API (200): no LobeHub-named thread above noise, grounding the thin-footprint observation

agent-platform on tomrochette.com

AutoGPT

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

LobeHub

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

agent-platform on tomrochette.com

AutoGPT

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

LobeHub

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#