Agenta
Open-source workspace for AI coworkers you run and automate from Slack or Telegram
Agenta treats agents as shared team infrastructure rather than one engineer's terminal session, and the self-hosted edition is the honest headline: MIT license, unlimited users, unlimited agent runs, nothing metered. The Cloud plans aren't priced per seat either — $29/mo and $299/mo buy credits, concurrency and governance, not extra humans. The v0.121.2–v0.121.6 run of releases put agents into Slack and Telegram with channel-wide read/post/search, added a template marketplace and moved model support to Claude Opus 5.5, Claude Sonnet 5.5 and Grok. If you need production alerting and on-call incident response, look at a monitoring platform instead; that is not what this is built to do.
Verified 3d ago · liveness 85/100 · cite: rightaichoice.com/tools/agenta
- Teams that want one shared, versioned workspace for building and reviewing recurring agents together
- Product managers and domain experts who'd rather refine an agent in chat than edit config files
- Engineering teams running code review, triage or research agents on Claude Code or Codex harnesses
- Organizations that need agents on their own infrastructure under an MIT license with no run metering
- Teams whose main need is production monitoring with alerting and incident response
- Buyers who want a fully managed platform and no deployment or upgrade responsibilities
- Workflows where Slack, Telegram or WhatsApp can't be the agent's interface
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Agenta if you need a monitoring and incident-response platform with production alerting and on-call rotation, since Agenta is built to run and log agents rather than to page you when a service degrades.
Credits pay for Agenta's included models and for runtime — the time your agents run — so long browser sessions and frequent scheduled runs burn through a Starter plan's 2,900 monthly credits faster than a few short
Per-seat SaaS for agent teams commonly asks $20-$40 per user every month; Agenta charges $29/mo (Starter) or $299/mo (Business) for unlimited team members, so a ten-person team lands far below seat-based peers. The trade is credit metering plus $10 per extra 1,000 credits. Teams with their own infrastructure should compare the free MIT self-hosted edition against any managed competitor before paying for Cloud at all.
In short
Agenta — Open-source workspace for AI coworkers you run and automate from Slack or Telegram. Best for Teams that want one shared, versioned workspace for building and reviewing recurring agents together, Product managers and domain experts who'd rather refine an agent in chat than edit config files, Engineering teams running code review, triage or research agents on Claude Code or Codex harnesses. Free to start; paid plans from $29/mo.
No head-to-head comparison for Agenta yet — the alternatives below are the closest matches we list.
What's new in Agenta
Checked 6 days agoAcross the latest 10 updates: 4 feature updates, 2 changelog entries and 4 news mentions.
Agenta v0.121.6 adds Claude Opus 5.5, Claude Sonnet 5.5 and Grok
Agents can run on Claude Opus 5.5, Claude Sonnet 5.5 and xAI's Grok, with a direct xAI provider option. Cost tracking now prices cached tokens and cache writes per call.
Agenta v0.121.6: create integrations, skills and automations with AI
Integrations, Skills and Automations now have a Create with AI option that sets them up via chat. Template-created agents prompt for required accounts before their first task.
Agenta v0.121.5 adds agent templates and a template marketplace
Users can start from ready-made agents with bundled skills, files and connections, or export a session as a template (.zip) for others to reuse.
Agenta v0.121.3 lets agents post, read and search Slack and Telegram channels
Agents connected to Slack or Telegram can act on those channels in any run, not just message replies, and can share a link to their current chat.
Agenta v0.121.2 adds Telegram agent publishing
Agents can be published to Telegram via QR code from the Publish button. Chats keep memory across messages.
Agenta v0.121.2 adds Slack agent publishing
Agents can be added to a Slack workspace from the Publish button and mentioned in any channel or thread; they reply in-thread, retain context, and mark progress with reactions.
Agenta publishes comparison of 9 AI coworker platforms for business teams
Agenta compares nine AI coworker platforms on recurring work, team sharing, integrations, model choice, deployment, control and price.
Agenta publishes comparison of open-source Claude Cowork alternatives
Agenta compares open-source alternatives for personal use, team collaboration, self-hosting, model choice, permissions and recurring work.
Agenta publishes list of Grok bot alternatives for 2026
Agenta's blog posts a 2026 roundup of alternatives to Grok bots.
Agenta publishes list of Manus alternatives for 2026
Agenta's blog posts a 2026 roundup of alternatives to Manus.
What people actually say about Agenta — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
96 mentions across 6 sources (Hacker News, YouTube, Product Hunt, Bluesky, GitHub, Lemmy) · researched Jul 25, 2026.
Average across the 6 sources that answered — each source counts once, not each post.
- +Unified playground for side-by-side model comparison across providers.
- +Live evaluation on edit speeds up prompt iteration significantly.
- +UI designed for non-technical experts to edit prompts collaboratively.
- +Open-source with self-hosting option avoids vendor lock-in.
- +Automated evaluation using LLM-as-a-judge or custom code.
- −Missing built-in approval workflows for enterprise governance.
- −Audit trail features are not yet available out of the box.
- −Scalability for high-throughput async LLM calls is unclear.
- −Trace sampling may drop edge cases under production load.
- −Self-hosted version lacks official support SLA.
- • Self-hosting may require DevOps effort and infrastructure costs
- • Higher-tier features like SSO are only on Enterprise plan
Viability Score
How well maintained and how widely used is Agenta? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Build and use AI coworkers through chat instead of config files
- Create integrations, skills and automations with AI via chat
- Publish agents to Slack from the Publish button; they reply in-thread with context
- Publish agents to Telegram via QR code with memory across messages
- Agents read, post to and search connected Slack and Telegram channels in any run
- Start from pre-built agent templates or export a session as a .zip template
- Reusable skills that carry agent know-how between runs
- MCP server connections for tools like Notion, GitHub, HubSpot and Linear
- Run agents on Claude Code, Codex or pi.dev harnesses
- Bring your own ChatGPT or Claude subscription, OpenRouter, Gemini, xAI or Ollama
- Run agents on Claude Opus 5.5, Claude Sonnet 5.5 and Grok
- Built-in browser plus persistent AI Drive folder of files per agent
- Scheduled runs, event triggers and webhooks with every run logged
- Cost tracking that prices cached tokens and cache writes per call
- LLM-as-a-judge evaluations, human annotation queues and trace-to-test-set conversion
About Agenta
Agenta is an open-source workspace for building AI coworkers — agents that carry out recurring team work rather than answer one-off prompts. You describe the job in a browser chat and the agent gets its own computer: a browser to research with and a persistent AI Drive folder to save files into. One more message puts it on a schedule or an event trigger, and every run is logged with tokens, latency and estimated cost. Agents are composable by design. They run on Claude Code, Codex or pi.dev harnesses and reach models through a ChatGPT or Claude subscription you already pay for, an OpenRouter or Gemini key, xAI, or a self-hosted Ollama model. v0.121.6 added Claude Opus 5.5, Claude Sonnet 5.5 and Grok as runnable models, with cost tracking that now prices cached tokens and cache writes per call. Skills carry the know-how, MCP servers supply the tools, and either can be swapped without rebuilding the agent — and since v0.121.6 both, plus automations, can be set up by chatting with an AI assistant. Conversation surfaces are where this diverges from a dashboard-first tool. Publish an agent to Slack from the Publish button, or to Telegram by QR code, and it keeps memory across messages. Since v0.121.3 agents can read, post to and search connected Slack and Telegram channels in any run, not just reply in the thread they were started from. Tool permissions sit behind an approval gate, and credentials live in a vault the agent can use but not read. Quality questions get answered in the same interface: a playground for side-by-side comparison, LLM-as-a-judge scoring, human annotation queues and trace-to-test-set conversion. Version history covers agent and prompt configurations. v0.121.5 added agent templates — start from a ready-made agent with bundled skills, files and connections, or export a session as a .zip your colleagues can reuse. The self-hosted edition ships MIT-licensed on Docker Compose with unlimited users and runs, so the Cloud plans are effectively a hosting and governance layer on top of it.
Behind the Verdict
Agenta's pitch is narrow and specific: a coworker, not a chatbot. The mechanics that back that up are the browser plus the per-agent AI Drive folder. An agent can go find something, then write the finished work into a shared folder your team reads — the homepage walkthrough has an agent query a PostHog funnel, spot a 4% week-over-week drop at step 3, and update a PRD file in the same run. That loop of research, create, save is what separates a coworker from a prompt window. The second differentiator is where the agent lives. Publishing to Slack happens from a Publish button; Telegram uses a QR code. Since v0.121.3 a connected agent can read, post to and search those channels in any run, not only answer the thread that started it. For a team already working in Slack, that removes a whole class of "go check the dashboard" behaviour. Model flexibility is real rather than decorative. You can point an agent at Claude Code, Codex or pi.dev, and back it with your own ChatGPT or Claude subscription, an OpenRouter or Gemini key, xAI, or a self-hosted Ollama model. v0.121.6 added Claude Opus 5.5, Claude Sonnet 5.5 and Grok, and cost tracking now prices cached tokens and cache writes so your per-run estimates don't quietly undercount. The self-hosted edition is MIT-licensed on Docker Compose, includes role-based access control and SSO, and imposes no metering on users, agents or runs — for a security team that already has Kubernetes-adjacent habits, that is the whole ballgame. Where it gets less comfortable: Cloud is metered by credits, not seats. Hobby gives you 500 bonus credits on signup and 75 free credits a day on up to 10 days a month, capped at 2 concurrent tasks, 1 project and 2 team members. Starter is $29/month for 2,900 credits and 20 concurrent tasks; Business is $299/month for 32,000 credits and 50 concurrent tasks, plus SSO, role-based access control, the SOC 2 Type II report and priority support. Extra credits run 1,000 for $10. Long browser sessions and busy scheduled automations are exactly the kind of usage that consumes credits quickly, so budget for the ceiling rather than the sticker. Audit logs, custom data retention, custom domains and a custom SLA are Enterprise-only, and on self-hosted those are Enterprise-tier items too — the free open-source build leaves deployment and upgrades to you. Build-through-chat is the design choice most likely to divide a team. A domain expert can refine an agent by talking to it, and v0.121.6 lets you create integrations, skills and automations the same way — but engineers who want a config file and a diff will find the primary interaction unfamiliar. The version history and trace-to-test-set conversion partly compensate, giving you a review trail even when the author never opened an editor. Compared with dashboard-first eval tools, Agenta's strengths sit in durable, recurring work with a shared file trail; compared with a platform like Manus, the differentiators are self-hosting under MIT and
Researching Agenta? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Agenta actually fits — and what changes day-one when you adopt it.
On Monday morning they message the SEO coworker in Slack to research the next best-tools article; the agent browses, saves the search-intent brief to its AI Drive folder and posts the outline back in-thread.
Outcome: The brief is reviewed and approved in Slack before the agent drafts it, so the article ships from a single conversation instead of a handoff chain.
They start from the code review agent template, wire it to GitHub and Linear on the Claude Code harness, and trigger it on every pull request; the AGENTS.md instructions escalate to the code owner when a diff touches auth, billing or migrations.
Outcome: Reviews arrive as inline comments with minimal suggested fixes, and the diff-touching-auth cases still land with a human owner.
They build a competitor research agent by chatting with it, connect it to Google Drive and Notion, and schedule it to run weekly with an event trigger on new signups.
Outcome: The agent writes each research brief into the shared workspace folder, and the per-run log shows tokens, latency and estimated cost so the credit budget stays visible.
Use Cases
- Set up a code review agent that reviews every pull request, leaves inline comments and tags the code owner when a diff touches auth, billing or migrations.
- Schedule a competitor research automation that runs every Monday and writes the brief into the shared workspace.
- Build an onboarding facilitator agent that queries PostHog funnels and updates the onboarding PRD with the results.
- Trigger a sales prospecting agent when a new signup occurs and have it research the prospect before outreach.
- Draft SEO articles end to end, from search intent and outline through to a prepared pull request.
- Self-host Agenta to keep LLM data and model credentials inside your own infrastructure.
- Run agent and prompt evaluations in CI with LLM-as-a-judge gates before a configuration reaches production.
- Review and approve agent runs from a phone when a scheduled automation flags an anomaly.
Models Under the Hood
as of 2026-10-16
Limitations
- Agenta Cloud meters usage by credits rather than seats.
- Hobby is $0 with 500 bonus credits on signup, 75 free credits a day on up to 10 days a month, 2 concurrent tasks, 1 project and 2 team members.
- Starter is $29/month with 2,900 credits and 20 concurrent tasks; Business is $299/month with 32,000 credits and 50 concurrent tasks.
- Extra credits cost 1,000 for $10.
- Governance features such as SSO, role-based access control, the SOC 2 Type II report and priority support sit on Business and above, and audit logs, custom data retention, custom domains and a dedicated SLA are Enterprise-only.
- Agents are built and run primarily through chat rather than config files, which suits domain experts but gives engineers no file-and-diff workflow.
- On the self-hosted side the free open-source edition allows unlimited users, agents and runs but leaves Docker Compose deployment, upgrades and operations to your team.
as of 2026-10-19
Verification history
We have re-verified Agenta 21 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 21 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Agenta tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Hobby (Cloud)
$0 forever
Ideal for
An individual testing whether a recurring agent is worth it, with 500 bonus credits on signup and 2 concurrent tasks, on 1 project with 2 team members.
What this tier adds
Free entry point: 75 free credits a day on up to 10 days a month, agents with their own browser and AI Drive, and Slack/Telegram publishing, but no monthly credit allowance.
Starter (Cloud)
$29/mo
Ideal for
A product or marketing team running agents in daily work, with unlimited projects, unlimited team members and 20 concurrent tasks.
What this tier adds
Adds 2,900 monthly credits (extras at 1,000 for $10), unlimited team members at no per-seat price, unlimited schedules and event triggers, shared credentials and live chat support.
Business (Cloud)
$299/mo
Ideal for
A team needing governance and priority support, such as a regulated buyer that must show SSO and a SOC 2 Type II report before rollout.
What this tier adds
Raises the monthly allowance to 32,000 credits and 50 concurrent tasks and adds SSO, role-based access control, the SOC 2 Type II report and priority support in a private Slack channel.
Enterprise (Cloud)
Custom
Ideal for
Organizations that need custom limits, security and legal review, and a contractual service level on Agenta-hosted infrastructure.
What this tier adds
Adds custom credits and limits, custom data retention and audit logs, custom domains, custom security and legal terms, and deployment support with a dedicated SLA.
Open source (self-hosted)
Free forever
Ideal for
Engineering teams that must keep LLM data and model credentials inside their own infrastructure, with no metering on users, agents or runs.
What this tier adds
Free MIT-licensed starting tier on Docker Compose with unlimited users, projects, agents, runs and evaluations, plus role-based access control and SSO — you supply the models and operations.
Enterprise (self-hosted)
Custom
Ideal for
Enterprises running the open-source edition in production that want commercial support, audit trails and a contractual service level.
What this tier adds
Adds audit logs, custom domains, deployment and upgrade support, security and architecture reviews, priority support in a private Slack Connect channel and a dedicated SLA.
Where the pricing makes sense
The company stage and team size where Agenta's pricing actually pencils out — and where peers do it cheaper.
Per-seat SaaS for agent teams commonly asks $20-$40 per user every month; Agenta charges $29/mo (Starter) or $299/mo (Business) for unlimited team members, so a ten-person team lands far below seat-based peers. The trade is credit metering plus $10 per extra 1,000 credits. Teams with their own infrastructure should compare the free MIT self-hosted edition against any managed competitor before paying for Cloud at all.
Setup time & first value
How long it actually takes to get something useful out of Agenta — broken out by persona, not the marketing-page minute.
Hobby: minutes to a working agent if you describe the job in chat and accept the 2-concurrent-task limit. Starter: most of a first day to wire Slack publish, approvals, shared credentials and the first schedule. Self-hosted: plan a day or more for Docker Compose, model keys and role-based access control before the team moves real work onto it.
Switching to or from Agenta
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Manus: rebuild the recurring job as an agent in chat, attach your own ChatGPT or Claude subscription or an OpenRouter key, then publish it to Slack or Telegram instead of a dashboard.
- →From Grok bot style workflows: start from the pre-built agent templates, then add Grok as the model via the xAI provider option and keep the automation on a schedule.
- →From a DIY Claude Code script: run the agent on the same Claude Code harness inside Agenta to gain run logs, version history, approvals and scheduled triggers.
- →From a hosted eval platform: move prompt and agent configurations into version history and use trace-to-test-set conversion to rebuild existing test sets from runs.
- ↗To a production monitoring stack: export run logs and usage analytics to a tool with alerting and on-call rotation, since Agenta does not handle incident response.
- ↗To a fully managed agent platform: replace Cloud with a vendor that absorbs upgrade and deployment work — note the custom terms, data retention and SLA would have to be renegotiated.
- ↗To raw harness scripts: take your AGENTS.md instructions, skills and MCP server list out of the workspace and run them directly on Claude Code or Codex.
Integrations
Resources & Guides
- Documentationagenta.ai
What is Agenta? - Docs
Agenta is an open-source LLMOps platform: prompt playground, prompt management, LLM evaluation, and LLM Observability all in one place.
- Resourceagenta.ai
Prompt Management, Evaluation, and Observability for LLM apps
Agenta is an open-source platform for building robust LLM Application. It provides tools for prompt engineering, evaluation, debugging, and monitoring of complex LLM Apps.
Tutorials & Learning
YouTube returned 6 videos for “Agenta”, and we withheld 6: 6 could not be judged, because “Agenta” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Agenta.
Official links
Tools that pair well with Agenta
Common stack mates teams adopt alongside Agenta, with the specific reason each pairing earns its keep.
Arize Phoenix
Open-source LLM observability and agent evaluation you can trace, score, and self-host on your own infrastructure.
OpenAgents
Open-source research platform (COLM 2024) shipping three language agents — Data, Plugins and Web — in one chat UI, for free.
Phoenix
Open-source LLM observability and agent evaluation you can trace, self-host, and eval end to end.
Alternatives to Agenta
View all
Arize Phoenix
Open-source LLM observability and agent evaluation you can trace, score, and self-host on your own infrastructure.
OpenAgents
Open-source research platform (COLM 2024) shipping three language agents — Data, Plugins and Web — in one chat UI, for free.
Frequently Asked Questions
Used Agenta? Help shape our editorial sentiment research.