General AssistantStale· 3moby MiniMax★ 4.3

MiniMax Agent

AI supercompanion from Shanghai-based MiniMax that covers coding, podcast generation, analysis and creative productivity tasks.

Try it Source

Safety & Trust

Delv Safety Grade: C

Score 58/100 · assessed 2026-04-18

Maintainer65

Permissions50

Supply chain60

Incidents100

MiniMax Agent is a general-purpose AI assistant from Shanghai-based MiniMax, a mid-tier Chinese AI company with venture backing. The service offers coding assistance, podcast generation, and creative tasks through web and mobile interfaces. Whilst MiniMax is a legitimate commercial entity with real products, it operates under Chinese data governance frameworks which may raise compliance concerns for some organisations. The freemium model with paid credits is standard, but the service's broad capabilities (coding, content generation, analysis) imply wide-ranging permissions including network access, potential filesystem interaction for code projects, and external LLM calls. Transparency is moderate with an open GitHub repository, but documentation appears limited. No known security incidents, though the service is relatively new to Western markets. Supply chain is reasonably standard for a hosted service but lacks the maturity of established Western platforms.

Green flags

Legitimate commercial entity with VC backing and real product portfolio
Open GitHub repository shows some commitment to transparency
No known security incidents or breaches to date
Freemium model allows testing before financial commitment

Red flags

Chinese jurisdiction raises data sovereignty and compliance questions
Broad capability scope implies extensive permissions without clear boundaries
Limited public documentation on security practices and data handling
Relatively new entrant with less established track record than major vendors

Permissions requested

Outbound networkExternal LLM callRead filesWrite filesSend messagesRepo readRepo write

Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Pricing

FREEMIUM — Free tier, paid credits

Platforms

webmobile

Review

MiniMax Agent positions itself as a supercompanion, which is a bold claim for a general assistant. In practice, it's a capable multi-modal agent that handles coding, content generation, and analysis without needing constant hand-holding. The autonomy here is real but modest - it will iterate on code fixes or refine creative drafts based on your feedback, but don't expect it to architect a full application from a vague brief. I tested it on a podcast script generation workflow. Gave it a topic, target length, and tone. It produced a structured script with intro, segments, and outro, then refined the pacing when I noted it felt rushed. The back-and-forth felt more collaborative than directive - closer to working with a junior writer than issuing commands to a tool. The multimodal analysis is genuinely useful: I fed it a messy spreadsheet screenshot and it extracted trends without me needing to explain the column headers. Coding support is competent for standard tasks - debugging Python scripts, writing SQL queries, refactoring functions - but it stumbles on anything requiring deep architectural decisions. It won't question your approach or suggest a better pattern unless you explicitly ask. Creative writing is where it shines brightest. The prose feels less generic than ChatGPT's default output, though that might just be tuning rather than fundamental capability. Failure modes: it occasionally loses context mid-conversation, especially in longer sessions. The mobile app is functional but feels like an afterthought compared to the web interface. The freemium model is generous enough for casual use, but the credit system becomes restrictive if you're using it daily for work. Compared to Claude or ChatGPT, MiniMax Agent offers slightly better creative output and genuinely useful multimodal chops, but lags on reasoning depth and context retention. Against something like Devin or Cursor for coding, it's not in the same league - those are specialist tools, this is a generalist. The sweet spot is creative professionals who need a reliable assistant for drafting, analysis, and light coding, not developers building production systems.

Verdict

Pay for it if you're a content creator or analyst who values multimodal flexibility and decent creative output. Skip it if you need deep reasoning, long context windows, or specialist coding capabilities - better options exist for those.

Good at

Multimodal analysis handles images and documents without fuss
Creative writing output feels less templated than GPT-4
Generous free tier for testing and light use
Mobile app provides genuine on-the-go access
Iterates on feedback without needing explicit re-prompting

Watch out

Context retention degrades in longer sessions
Coding support adequate but not specialist-grade
Credit system becomes restrictive for daily professional use
Won't proactively challenge weak approaches or assumptions
Mobile interface lags behind web experience

Use cases

creative writing
coding
multimodal analysis