Open Bee
Get started

What's next

Roadmap

A 12-month view of what we're building. Dates are estimates β€” we publish them anyway because moving the goalposts in private is worse than missing them in public.

Want to influence priorities? Vote and comment on the public board at GitHub Discussions. The most-upvoted items in Next get pulled forward.

Shipping now

Now (in progress)

Β· Jun 2026
  • Landing & marketing refresh

    Repositioning the whole site around the prompting-agent platform, in English / Bahasa Indonesia / δΈ­ζ–‡.

  • Agent run trace persistence

    Store per-step outputs on each run so the runs view can replay a multi-step agent end to end.

  • Recipe gallery groundwork

    Curated, one-click pre-built agents β€” the same idea as the prompt template library, but for whole agents.

Next

Next quarter

Β· Q3 2026
  • Tool use in agents

    Let agent steps call tools β€” web search, HTTP, your own functions β€” so agents can fetch context, not just transform it.

  • Agent recipe gallery

    One-click, pre-built agents for common jobs (ticket triage, code review, doc writer) with typed inputs.

  • SOC 2 Type I report

    With Vanta. Available under NDA from /security once we have the report.

  • Audit log forwarding

    Stream audit events to your SIEM (Datadog, Splunk) for customers with their own retention policies.

Later

Later this year

Β· Q4 2026
  • Team prompt + agent sharing

    Org-level libraries, review/approve flows, and per-agent access by role.

  • EU data residency option

    Frankfurt cluster, gated by enough EU customers to justify the operational cost.

  • Customer-managed encryption keys (CMEK)

    For enterprise customers who want to hold the key that decrypts their BYOK + integration secrets.

Shipped

Recently shipped

  • Dashboard & usage analytics

    Runs, tokens, cost, a 14-day trend, top agents, and recent team activity across the workspace.

  • Prompt Studio

    Author / test / version prompts with a 17-template library, fillable variables, and one-click 'use in agent'.

  • n8n integration (two-way)

    Call agents from n8n via the public API, and have agents trigger n8n on completion (with a shared secret). Ready-made workflow templates in the docs.

  • Multi-step agents

    Chain steps where each output flows into the next via {{previous}} / {{stepN}}, with a per-step model + thinking level.

  • Agents + public API

    Build agents and call them from your systems: POST /api/v1/agents/:id/run with an API client key + burst/daily rate limits. BYOK provider keys per workspace.

  • AI core β€” model router + thinking levels

    Recommend the best model per task (accuracy / speed / cost), light / standard / deep reasoning, and a live cost estimate.

  • Roles + English-default i18n

    owner / admin / editor / viewer permissions; English-default UI with Bahasa Indonesia + δΈ­ζ–‡; auto-migrate on every deploy.

  • Open-source agent runtime

    Apache-2.0 daemon, plugin SDK, and reference tools you can self-host β€” code and files never leave your machine.

Roadmap Β· Open Bee AI