Agenta
agenta.ai
- Free plan
- Open source
- Web
- API
- Self-hosted
What Agenta is
Agenta is an open-source workspace for teams to build, use, and automate AI coworkers. Teams can work with agents in Slack or Telegram, start from templates for engineering, customer support, sales, company knowledge, or operations, and schedule runs or trigger them with events. Its monitoring tools list agent traces, execution logs and errors, version history, evaluations, and token usage with estimated model costs. Agenta names connections including Slack, Notion, GitHub, HubSpot, Google Drive, Linear, Zendesk, and Discord, and says it supports more than 1,000 integrations. Model providers include OpenAI, Anthropic, Gemini, Mistral, and others. The free Cloud Hobby plan allows two team members and 5,000 agent runs a month; runs stop at that limit. Pro costs $29/month and includes 10,000 runs, with extra runs charged separately. Model-provider costs are separate. Self-hosting puts operation of the app, storage, runner, upgrades, and backups in the user's infrastructure.
Who it suits
Agenta suits teams building and automating AI agents across work tools, and teams that need traces, evaluations, or model-provider choice. Self-hosting may suit teams that want to control data location and operate the application in their own infrastructure.
Good to know, and what to watch
- Free Cloud Hobby plan allows 5,000 runs monthly
- Supports scheduled and event-triggered agent runs
- Lists more than 1,000 integrations
- Self-hosting lets users control data storage location
- Business plan includes RBAC and SSO
- Free Hobby runs stop at 5,000 per month
- Model-provider costs are paid separately
- Pro costs $29/month
- Hobby and Pro list community support
Plans
Every plan and priceEnterprise
Enterprise (Cloud)
Enterprise (self-hosted)
Enterprise self-hosted
Free plan
Getting started
- Open Agenta’s website
- Choose the Hobby, Pro, Business, Enterprise, or self-hosted option
- Create a workspace and connect your own model-provider credentials
- Select an agent template or build an agent
- Connect services such as Slack, Notion, GitHub, or other listed integrations
- Set schedules or event triggers for runs
- Use the web, API, self-hosted deployment, or mobile app
Limits to check
The free Cloud Hobby plan includes two team members, 5,000 agent runs per month, 20 evaluations per month, and one-week trace retention; runs stop at the monthly run limit. Pro costs $29 a month with 10,000 included runs and charges 5 USD for each additional 10,000 runs.
The Club's review
Agenta is an open-source workspace for creating, running, and automating AI coworkers. It is best suited to teams that want agents connected to workplace software, with developers able to choose cloud hosting or operate the stack themselves. Its unusually broad agent and observability scope is attractive, but usage quotas and provider bills make careful cost planning essential.
Overview
Agenta brings agent creation, workflow automation, and operational visibility into one workspace. Pre-built templates cover engineering, customer support, sales, company knowledge, and operations, giving teams a practical starting point instead of an entirely blank canvas. Agents can be used through Slack or Telegram, while connections to Notion, GitHub, HubSpot, Google Drive, Linear, Zendesk, and Discord—and more than 1,000 integrations overall—make it relevant to cross-tool business processes.
Scheduled and event-triggered runs support recurring jobs and responsive workflows. Traces, execution logs and errors, version history, evaluations, and token usage with estimated model costs provide the information needed to diagnose failures and compare changes. Agenta works with OpenAI, Anthropic, Gemini, Mistral, Groq, MiniMax, Together AI, OpenRouter, Azure OpenAI, AWS Bedrock, Google Vertex AI, and OpenAI-compatible endpoints. You supply credentials and pay model providers directly, so subscription pricing does not represent the whole bill.
Key features
Building, testing, and version control
Prompt templates, prompt testing, prompt versioning, and API or SDK access suit teams moving from experiments toward repeatable agent development. Evaluations are useful for checking behavior as prompts and workflows change, while version history provides a record of those changes.
Automation and observability
Schedules and event triggers let agents operate without a person starting every run. Run traces, errors, execution logs, and estimated token costs help an engineering or operations team see what happened and where spending is going. The free cloud tier's small evaluation allowance, however, limits how extensively a team can assess agents before upgrading.
Security and administration
Business adds role-based access control, single sign-on, and a SOC 2 Type II report, making it the first cloud tier aimed at organizations with formal access and assurance requirements. Enterprise adds audit logs, custom domains, dedicated support, custom service levels, and tailored security and legal terms. Those controls are not part of Hobby or Pro.
Self-hosting and mobile access
The open-source option lets an organization run the application, storage, agent runner, upgrades, and backups in its own infrastructure, retaining control over data location. That freedom comes with operational responsibility. A mobile app supports chatting with agents, continuing conversations, approving tools, and managing workspaces from a phone, extending access beyond the desktop.
Pricing
| Plan | Price | What you get |
|---|---|---|
| Hobby | Free, billed $0 forever | 2 team members, 5,000 agent runs/month, 20 evaluations/month, 1-week trace retention, community support |
| Open source | Free, billed Free forever | Self-hosted; unlimited users, projects, agents, workflows, schedules, and event triggers; self-managed trace retention; bring your own models and credentials |
| Pro | $29 a month, billed $29/month | Unlimited team members, 10,000 runs/month, $5 per additional 10,000 runs, unlimited evaluations, 1-month trace retention, community support |
| Business | $299 a month, billed $299/month | 10,000 runs/month, $5 per additional 10,000 runs, RBAC, SSO, SOC 2 Type II report, 3-month trace retention, priority support and private Slack Connect |
| Enterprise | Custom pricing | Cloud or self-hosted deployment, custom usage and retention (cloud), audit logs, custom domains, dedicated support, custom SLA; self-hosted includes deployment and upgrade support |
Hobby is a sensible trial for a two-person team, but runs stop at 5,000 per month and traces last only a week. Pro removes the seat ceiling and evaluation limit, fitting active teams that can stay within 10,000 runs or accept the overage charge. Business is justified when identity controls, compliance reporting, longer retention, and priority assistance matter; it costs substantially more without increasing the included run allowance. Open source is the strongest value for organizations prepared to operate infrastructure, while Enterprise is for bespoke governance and support. Every cloud plan leaves model-provider charges separate.
Platforms
Agenta is available through its web interface, API, and self-hosted deployment. The mobile app adds phone-based agent conversations, approvals, and workspace administration. Self-hosting maximizes control over storage and operations; cloud hosting reduces that operational burden.
Who it's for
Choose Agenta if your team wants departmental agent templates, broad workplace integrations, automated runs, and detailed run oversight in one system. It is particularly suitable for teams that need Slack or Telegram access, multi-provider model choice, or a self-hosted route. It is less suitable when workloads routinely exceed cloud quotas, when you want provider costs bundled into one subscription, or when you cannot take on self-hosting responsibilities.
Pros and cons
- Pros: Broad integrations and department-focused templates shorten the path to useful workflows.
- Pros: Traces, logs, evaluations, version history, and cost estimates support serious operational review.
- Pros: Self-hosting offers unlimited core workspace resources and control over data location.
- Pros: Multi-provider support avoids locking a team to one model vendor.
- Cons: Hobby stops runs at 5,000 per month and limits teams to two members.
- Cons: Cloud plans require separate provider credentials and direct model payments.
- Cons: Advanced access, compliance, and support features begin at the $299 Business tier.
- Cons: Self-hosted users must manage deployment, storage, runners, upgrades, and backups.
Alternatives
For prompt-focused gateway management with Android and iOS apps, a one-seat free tier, and a 10,000-call hard cap, choose PromptHelm. Maxim AI is an alternative with a free developer plan for up to three seats, one workspace, 10,000 logs per month, and three-day retention. Choose Vellum if you prefer a broad platform lineup and pay-as-you-go credits on its free entry plan. Promptfoo suits teams prioritizing red-team probes and LLM evaluation with a local or self-hosted community edition. ChainForge is a free, open-source, self-hosted and web alternative. Langfuse is a monitoring-oriented option with a free 50,000-units-per-month hobby tier and 30-day data access. Prompt Manage is aimed at smaller prompt collections and run volumes, while PromptHub offers unlimited public prompts and 2,000 requests per month on its free tier.
For category browsing, see AI Prompt Management Software, AI Prompt Generators, AI LLM Observability Tools, Prompt Management Software, and AI Agent Observability Tools.
Verdict
Agenta is a strong choice for teams building multi-step AI coworkers across business tools and wanting both cloud convenience and a self-hosted escape hatch. Its combination of templates, automation, multi-provider support, and run-level oversight is the main reason to choose it. Look elsewhere if your priority is an all-inclusive model bill, very high cloud volume without quota management, or advanced governance at a lower entry price.
The verdict
Agenta offers cloud and self-hosted paths for teams building AI coworkers, with run limits and model charges to account for. Compare the free plan's cap with your expected usage, and remember that provider costs are separate.
Questions people ask
Is Agenta open source?
Yes. The Open source plan is free forever and self-hosted, with unlimited users, projects, agents, workflows, schedules, and event triggers.
What does the free cloud plan include?
Hobby is free forever for two team members, 5,000 agent runs per month, 20 evaluations per month, and one-week trace retention.
How much is Pro?
Pro costs $29 a month and includes unlimited team members, 10,000 agent runs per month, unlimited evaluations, and one-month trace retention. Additional 10,000 runs cost 5 USD.
Are model costs included?
No. Users connect their own provider credentials and pay model providers directly.
Which deployment options and platforms are available?
Agenta supports API, self-hosted, and web access. Its mobile app supports agent chats, tool approvals, and workspace management.
What security controls are listed?
The Business cloud plan includes role-based access control, SSO, and a SOC 2 Type II report. Enterprise adds audit logs and custom security and legal terms.
What the maker publishes
- Agent creation
- The homepage presents pre-built agent templates for engineering, customer support, sales, company knowledge, and operations. source
- Agents API
- An HTTP Agents API accepts stored or inline configurations and supports JSON, SSE and NDJSON responses with resumable sessions. source
- Automation
- Agenta supports scheduled and event-triggered agent runs. source
- Background automation
- Agents can run on schedules or be triggered by events from connected applications. source
- Company facts on LinkedIn
- LinkedIn lists Agenta as privately held, founded in 2023, headquartered in Berlin and employing 2–10 people. source
- Company location
- Agenta's imprint identifies Agentatech UG and gives its address as c/o betahaus, Rudi-Dutschke-Straße 23, 10969 Berlin, Germany. source
- Custom providers
- OpenAI-compatible endpoints, Ollama, vLLM, LocalAI and self-hosted models can be configured with a base URL and credentials. source
- Data deletion
- The terms state that customer data is deleted from production systems within 30 days after service termination, with backups retained up to 180 days. source
- Data regions
- Agenta Cloud offers US and EU hosts, while self-hosted customers choose their own infrastructure and data location. source
- Evaluation SDK
- The Python SDK supports evaluator creation and complete evaluation runs, including LLM-as-a-judge workflows. source
- Framework integrations
- Official integration guides cover Agno, DSPy, Google ADK, LangChain, LangGraph, LlamaIndex, OpenAI Agents SDK, PydanticAI and Vercel AI SDK. source
- Funding investors
- Crunchbase lists InReach Ventures and Antler as Agenta investors. source
- GitHub adoption
- The public repository page showed about 4.8k stars and 686 forks when opened. source
- Headquarters
- Berlin, Germany source
- Independent category listing
- G2 places Agenta in its Prompt Management Tools category and presents LaunchDarkly, TextExpander, LangChain, Arize AI and PromptLayer as alternatives. source
- Independent funding report
- Oryndex reports a disclosed $108,000 seed round and an additional undisclosed Antler investment in early 2024. source
- Integrations
- The homepage names Slack, Notion, GitHub, HubSpot, Google Drive, Linear, Zendesk, and Discord among its connections and states that it supports more than 1,000 integrations. source
- Integrations
- The documentation describes more than 1,000 integrations through built-in connectors, MCP servers and Composio. source
- Latest changelog
- A September 30, 2026 release added Claude Opus 5.5, Claude Sonnet 5.5, xAI Grok support and revised token-cost accounting. source
- Legal entity
- The terms identify the company as Agentatech UG (haftungsbeschränkt), registered at Rudi-Dutschke-Str. 23, Berlin, Germany. source
- Licence
- The public GitHub repository describes Agenta as open source under the MIT License. source
- Limit: code tools
- The self-hosted runner disables custom code tools as a security decision, so agents cannot execute arbitrary shell-based code tools there. source
- Limit: proxy mode
- The quick-start documentation says proxy mode does not support streaming responses and adds approximately 0.3 seconds of latency. source
- Limit: third-party app tools
- Self-hosted app tools and event triggers require a Composio API key; without it those capabilities return 404. source
Checked against the maker's pages on 14 Oct 2026.