Roadmap AgentZ ships features often, and users file requests for more. The lists below come from the public issues, releases and announcements in the accuknox/agentZ repository. An open issue is a request. It is not a commitment, and a request can change or close.
Recently Shipped Each row links to the release or announcement for the feature. The latest release is v0.28.1 . Its notes list container images and a changelog link only.
Status Feature Source Date Shipped Custom tools. Add a Bash, Python or Node.js script as a tool that an agent can call. See Custom tools . v0.28.0 and announcement 2026-10-05 Shipped Workflow evaluations (preview). Run one workflow on several models, let a judge model score each run and compare the results. See Evaluations . v0.25.0 and announcement 2026-09-30 Shipped Remembered dropdown choice. The web app remembers the last option you selected in a dropdown. v0.23.0 2026-09-17 Shipped Coding workspaces. A second workspace type next to General purpose. v0.22.0 2026-09-16 Shipped Grouped sidebar items. The sidebar lists the less-used items in a separate group. v0.21.0 2026-09-14 Shipped Sarvam AI provider. Sarvam AI is a built-in inference provider. See Use Sarvam AI as your model . Announcement 2026-09-08 Shipped Agent-built dashboards. Ask an agent for a view of your data. It picks the charts and builds the dashboard. See Dashboards . Announcement 2026-09-08 Shipped Organizations and workspaces. Sign-up creates an organization, and each team gets a workspace. Workspaces inherit shared sandboxes, skills, MCP connections and inference providers. See Organizations and workspaces . Announcement 2026-09-08
Open Requests by Theme Each row is an open GitHub issue, grouped by topic. Select an issue to read the full discussion.
Observability and Traces Status Issue Request Open #17 Search needed in Traces Search the hundreds of tool calls in a trace to find one specific call. Open #24 Visibility/Telemetry should be shown in the context of the workflow Show alerts, telemetry and traces inside the workflow view, with floating controls. Open #30 show timeseries graph based view for system events Show file, process, network and domain access on a time series graph, with blocked events highlighted. Open #51 Add Filters support for RunTime Telemetry Filter runtime telemetry by process, file path accessed, destination domain and action. Open #53 Include Model Reasoning in Traces Detailed View under Spans tab Show the model reasoning in the Spans tab, so you can improve your prompt. Open #54 User-friendly view for token usage and activity/event log Add one in-product view of token usage over time and an activity log of what used credits. Open #83 Connect external SIEM with AgentZ traces/telemetry Send model calls, tool calls, skill executions, runtime telemetry and workflow logs to an external SIEM.
Workflows MCP Connections and Setup Access, Keys and Quotas Skills and Agents Chat and Files Recently Closed These issues closed on or after 2026-08-21. GitHub records each one as completed, as a duplicate or as not planned. The last column shows which.
Status Issue Request Closed Closed #104 show workflow elapsed time Show the workflow execution time, the time budget and the time for each step. 2026-09-07, completed Closed #107 Local characters not accepted in chat input (Sarvam inference provider) Bug report: the chat input did not accept messages with the Sarvam AI provider. 2026-09-07, completed Closed #49 Network call count in Traces is wrong Bug report: the network call count in Traces was wrong, and the page needed pagination. 2026-08-28, completed Closed #47 Thought Process Dropdown for Agent closes when agent is working on a prompt Bug report: the thought process dropdown kept closing while the agent worked. 2026-08-26, completed Closed #22 adding organization/tenant support Make the platform multi-tenant. At the time, it supported many users in one tenant. 2026-08-21, completed Closed #55 Prompt Navigation is monotonous can we have a floating prompt index Add a floating prompt index to move between prompts. 2026-08-21, completed Closed #61 Failed runs don't say what actually blocked them Name the real blocker of a failed run, such as a blocked domain, a missing binary or an unattached secret. 2026-08-21, completed Closed #62 MCP reports a connection as successful even when it doesn't work Report an MCP connection as successful only when a real call proves it works. 2026-08-21, completed Closed #66 Config agent: diagnose a blocked run and offer to fix it, with explicit per-run consent Let an agent diagnose a blocked run and offer a fix, with consent for that run only. 2026-08-21, duplicate Closed #71 Missing sandbox packages require a manual detour instead of an in-chat install offer Offer to install a missing sandbox package from the chat. 2026-08-21, duplicate Closed #72 Escalating an MCP tool from read to write access means abandoning the workflow Ask in the chat for write access to one MCP tool, for one run only. 2026-08-21, duplicate Closed #70 Secret naming conventions (like GITHUB_TOKEN) are undocumented tribal knowledge Suggest the usual secret name, such as GITHUB_TOKEN, from the host in the secret form. 2026-08-21, not planned Closed #73 MCP tool lists have no grouping and won't scale to 100+ tool connectors Group the tools of an MCP connection into categories. 2026-08-21, not planned Closed #75 The file tree and artifact preview are easy to miss entirely Make the file tree and the artifact preview easier to find. 2026-08-21, not planned
Request a Feature Search the open issues first. If your request exists, add your use case to it. If it does not exist, open a new issue and describe the problem you want to solve.