Releases
What is new in MaestroHub.
The highlights of each release, newest first. The Knowledge Graph is in development.
3.1.0
10 November 2026
LatestStore and forward holds your data through outages and says what happened to every write.
Store and forward, for every eligible connector
- One write, one outcomeEvery buffered write is delivered, held in its buffer, kept in Failed messages, or counted as lost with its reason. Outage reports add up. Read more
- Outages use no attemptsAn outage is held for as long as the disk has room, and the buffer delivers when the destination is back. Only a destination that answers with errors while it takes other messages counts against a message.
- Nothing dropped silentlyA full disk is reported as no room with the remedy, a store that can no longer be written is shown as down, and a write its connection cannot buffer is named at every start.
- A full disk recoversA full data volume recovers without a restart, and a frozen one stalls only the store: health, sign-in and reads keep answering.
- Unreachable, bufferingA connection in an outage whose outputs hold messages reads Unreachable, buffering instead of failed, with how many messages are held and how old the oldest is.
- Drop policyBuffer settings choose what happens when a buffer is full: Drop oldest, Reject new or Dead-letter new. Every change is audited.
- A pool that follows the diskUntil someone sets it, the buffer pool is sized from the message store's room instead of a fixed number.
- Changing an output keeps its bufferEditing an output's mode, delivery order, freshness or destination keeps its buffer taking writes.
App Studio and Demo Plant, Beta
- Apps on MaestroHub's own addressApp Studio apps are served from the address you already use, with no second host name or certificate, and each app still runs isolated.
- One go-live doorEvery way to make a version live goes through one check by a signed-in person, with one audit row naming the approver. An AI client cannot make a version live.
- Bindings by nameApps read every target from their manifest, so an app moves to another line or plant by its manifest.
- Several demo plantsThe Demo Plant becomes a catalog of plants, with a discrete-manufacturing plant beside the bottling plant.
3.0
5 November 2026
Every value explains itself, agents ask before they write, and upgrades are one command.
Trust in the data
- Lineage mapOpen any value and see where it came from: the inputs as they were at that moment, the pipeline run and the machine address, up to ten hops back. Read more
- Probable cause on alertsAn alert walks the lineage of the record that fired it and ranks the suspect inputs, bad quality first.
- Every writer namedValues written by a person, an API client or an AI agent carry who wrote them, and every stored value records the pipeline run and function version that produced it.
- Quality with a reasonEvery reading carries good, uncertain or bad from the device through each pipeline step into storage, with the reason when it is not good: wrong unit, missing samples, out of order, duplicate or late. Charts and reads can show good data only.
- A lost link is marked BadWhen a connection drops, the last value is recorded as Bad at the moment of loss, so a silent gap never looks healthy.
- Late and corrected valuesA reading replayed after an outage is stored as late and never overwrites the live value or drives a machine. Readings dated in the future are refused, and a correction keeps the original and names who made it.
- Device time and unitsOPC UA readings keep the device's own timestamp and engineering unit.
- Empty runs flaggedA pipeline run that reports success but delivered nothing is flagged for attention.
- History per topicEach topic keeps 1, 7, 14 or 32 days, 1 day by default, with every licence.
Alarms and notifications
- Shelving and alarm historySilence a known flapping alarm until a set time, find it under Shelved, and look back at its past alerts.
- Slack, Teams and webhooksAlarm fires and clears go to Slack, Teams or any webhook, filtered by severity and throttled, with a log of every delivery.
- EscalationA critical alarm nobody acknowledges or shelves in time is sent again, naming the action needed.
- Connection alertsA connection that keeps failing raises a notification within seconds and announces its recovery.
- Your notification feedA bell and a feed page tell each person about what concerns them: access granted or revoked, access requests, licence status and security alerts. Each type can be muted.
- Licence remindersAdministrators are reminded a month, a week, three days, a day and an hour before the licence expires.
Agents and governance
- Approval gateA pipeline can pause at an Approval node until a person answers, then carry on with that answer, with every step in the audit trail. Available as the Approvals add-on.
- Toolsets per AI clientAn AI client connected over MCP asks for only the tools it needs, such as pipelines or connections. Permissions still decide what it may do.
- Tokens scoped per workspaceA personal access token holds its own permissions for each workspace it names, and works nowhere else.
- Audit names the personWhen an agent or the system acts, the audit row names the person responsible. Exports to security tools in CEF and ECS carry the session and a unique event id.
- Stewards control writersA steward can close a branch of the namespace to other writers, and the share dialog shows who may write there. With Advanced Access Control, an add-on on any licence, and in the free trial.
Access control
- Request accessA person without access asks for it from the screen they are on, and the named steward for that data approves or denies. Grants can be time-limited or conditional. With Advanced Access Control, an add-on on any licence, and in the free trial.
- Permissions per branchGrant or deny access on any branch of the namespace, and it holds when a topic is renamed. With Advanced Access Control, an add-on on any licence, and in the free trial.
- My AccessEveryone can see their own roles, what is shared with them, and what they can and cannot do.
- Changes apply at onceWhen an admin changes or removes a permission, open browsers update straight away and the person is notified.
- Owners for everythingWorkspace admins can give an owner to anything that has none.
Delivery
- Delivery order per outputIn order for setpoints and commands, any order for timestamped readings. Read more
- Freshness limitsBuffered commands older than a limit you set are set aside for inspection, so a stale setpoint is never sent.
- Pause, hold and kill switchPausing can hold delivery and resume it in order, and the kill switch stops intake without losing what was already received.
- Disk floorA free-space floor protects the machine: it warns as the disk fills, refuses new values with a clear reason at the floor, and resumes on its own.
Connectors
- 111 connectorsUp from 47 in 2.6, from PLCs and robots to SAP HANA, ServiceNow, Splunk and the AWS, Azure and Google Cloud services.
- PLCs and machinesKeyence KV, Panasonic FP, Siemens LOGO! and S5, Opto 22, Beckhoff TwinCAT, FANUC FOCAS and Mitsubishi MELSEC FX3, read and written by their native addresses.
- Rockwell, completeEtherNet/IP adds Micro800, whole UDTs, automatic types and a tag picker that imports your L5X file.
- IEC 60870-5-104Monitor, interrogate and command substations and grid equipment, over TLS. A control command is never re-sent automatically.
- PI SystemConnect natively, or read current, recorded and interpolated values through the PI Web API and browse the Asset Framework. Read more
- SAP IDoc and Event MeshSend goods movements, deliveries and orders into SAP ECC and S/4HANA, parse IDocs SAP pushes out, and exchange events over SAP Event Mesh. Read more
- Snowflake StreamingRows land in seconds through Snowpipe Streaming, exactly once, with no warehouse running. Read more
- AWSDynamoDB, Redshift, Timestream, Athena, Kinesis, Data Firehose, Glue, EventBridge, IoT Core and IoT Greengrass.
- Microsoft AzureEvent Hubs, Service Bus, Cosmos DB with its change feed, Synapse, Stream Analytics and Azure Functions.
- Google CloudCloud Storage, Firestore, Bigtable, Dataflow, Cloud Functions and Vertex AI.
- Messaging and databasesAMQP 1.0, SOAP web services, Apache Cassandra, ODBC and Prometheus.
- Raw TCP and UDPReach devices that speak their own socket protocol, with send, receive and request and reply.
- Governed MQTTDevices publish into the namespace through one guarded entry point, MaestroHub's own built-in broker is the only MQTT broker and supports TLS, and revoking a credential ends its live sessions.
Pipelines and functions
- Faster buildingThe designer adds node search and recents, suggestions from real last-run outputs, Format, and warnings at save about missing references.
- Shared JavaScript libraryReusable functions live in one library per workspace, and each pipeline stays on the version it saved until someone moves it on.
- Tests beside the codeWrite and run tests next to a function and see a clear pass or fail.
- Quality on publishPublish steps can mark data good, uncertain or bad, and a pipeline can only lower the quality it received.
- HTML extractPull values out of web pages and HTML responses.
- Longer jobs, bigger queriesFunctions can run for up to an hour, SQL queries return far more rows, and cancelling a query stops it on the database too.
Operations
- Upgrade from 2.6Check, apply and restore from the command line. The Upgrades page lists each run with its backup, roles and grants come through unchanged, and a 2.6 Docker Compose install upgrades by changing the image tag.
- Versioned configurationConnections, functions and topics are versioned, so changes can be reviewed and rolled back.
- One binaryThe context-engine binary replaces the separate lite build.
- Single-node Helm chartInstall the whole platform on Kubernetes as one workload, against your own database and broker.
- Azure IoT Edge moduleDeploy MaestroHub as an IoT Edge module that uses the Edge hub for messaging.
- Overview with a verdictThe Overview opens with a status verdict and a map of connections and pipelines with uptime bars, names silent pipelines and clears a failure once it is fixed.
- Connection HealthOne page shows which connections are down, since when, how often they flap and the top reasons.
- Platform HealthThe health page opens with an overall verdict and names any service that is down with the next step.
- Subscriptions that healSee which pipelines need which device data, and dropped subscriptions are restored on their own.
- Clearer run historySkipped and throttled triggers are visible, and repeated identical failures collapse into one row.
- Move work between instancesConnections, functions, topics, pipelines and dashboards travel as bundles that rebind on import.
- WorkspacesAdmins see every workspace's seats, status and maintenance window on one page and can restore a deleted one. Maintenance pauses its pipelines and connections, and each workspace's admins read only their own logs. With more than one workspace.
- TimescaleDB retentionTimescaleDB installs get retention tiers and compression, and the topic page says whether the retention it shows is enforced.
- Behind a reverse proxyListen on a Unix socket, keep a base path, and put the data directory where you choose.
Security
- Keys unique to each installA new install creates its own keys at first boot, refuses to start on a publicly known key, and an admin tool re-encrypts existing data with a new one.
- More secrets encryptedSMTP and SSO passwords and cloud connector keys are encrypted at rest, including data upgraded from 2.6.
Interface
- Topics that explain themselvesA topic opens with what the signal is doing and why, its current value, one timeline of what went wrong, and stale signals marked in the tree.
- Bulk topic actionsSelect many topics to configure, share, assign stewards to or delete together. A deleted topic can be restored for seven days.
- Connection pickerA compact list with category filters and the popular connectors first.
- Recent items in searchThe search box opens with the pipelines, connections, functions, dashboards and runs you opened last.
- Errors that say what to doConnection and server errors explain what went wrong and the next step, instead of a status code.
- Help for operatorsHelp pages written for the plant floor, with explanations of terms where they appear.
- Favorites and dark themeFavorite pages sit in the top bar and follow you to any browser, and a dark theme is available.
- Dashboard variablesCustom chart panels can use dashboard variables.
Beta, by early access
- MaestroThe AI assistant inside MaestroHub, acting inside the permissions of the person using it.
- App StudioDescribe the app you need and it runs on your plant's data. Every app does only what the person using it is allowed to do.
2.6.5
August 2026
The foundation: connect, give values meaning, deliver without loss, and let your own AI client in.
Highlights
- Context EngineConnectors to machines and systems, one governed namespace, quality and origin on every value, pipelines.
- Store and forwardA buffer on disk for every destination, failed-message triage and an outage simulator. Read more
- MCP serverYour own AI client, such as Claude, works with MaestroHub inside a person's permissions. Read more
- Access and auditRole-based access, single sign-on over OIDC, SAML and LDAP, and a hash-chained audit trail.
Try the latest release.
The free trial runs the current release with every feature. The full release notes are in the documentation.