Planera, the San Jose scheduling platform that describes itself on its own homepage as “AI-NATIVE SCHEDULING SOFTWARE” where “CPM meets AI in a visual, collaborative platform,” published its November product update with a last-updated date of 9 November. The post, written by Matt Burness, the company’s head of product, runs to fourteen items. The post’s own header image names three of them as the AI additions: a Constructability Agent, a Planera and Procore agent, and Planera MCP.
Read together, they describe a company that has decided AI may compute over a schedule and change nothing in it.
Planera MCP
The headline for this one is “Get answers from your Planera schedule in Claude or ChatGPT.” The post says “The Planera MCP brings your live Planera schedules and Planera’s CPM logic into Claude and ChatGPT,” and that “It’s read-only and only sees what your Planera login can see.” The question types it names are “critical paths, version comparisons, what-if scenarios, and quality checks.”
The phrase worth stopping on is and Planera’s CPM logic. Most of the Model Context Protocol connectors this publication has covered in construction expose records, and operations on records: forms, documents, model elements, ledger entries. ProjectTeam’s server names the forms an assistant can read and write and the four actions it keeps in-platform. CMiC’s NexusAI made the permissions plumbing its headline feature. Nemetschek put a public MCP server under Archicad in beta and MCP support into Bluebeam Max at its new Max tier.
The nearest precedent for what Planera describes is Qonic, whose MCP server can calculate quantities as well as query and change elements — the platform performing a domain computation on the assistant’s behalf, rather than handing over data for the model to work out. Autom8Labs comes at the same gap from the other side, publishing 22 free MCP skills as a method layer on the argument that giving an agent access to an application is not the same as telling it how to work in one.
A critical path sits further along that line than a quantity does. It is not a fact to be fetched; it is the output of a calculation over a network of activities, durations, constraints and calendars, and a language model asked to work one out from a task list is doing a kind of work it is comparatively poorly suited to. Planera’s own guide to AI scheduling tools, last updated 6 November, states the design explicitly: “The connector runs Planera’s own scheduling engine,” so “a critical path or what-if answer matches what the team sees in Planera.”
That is the distinction between a connector that gives a model data and one that gives it a calculator. If the claim holds in practice, an answer in a chat window and an answer in the Gantt chart cannot drift apart, because there is only one engine producing both. This publication has not tested the connector and is reporting the company’s description of it.
The Constructability Agent
The second item sits under the heading “Catch sequencing issues that standard schedule checks miss.” It runs from the side panel of Planera’s Schedule Quality Report, and the post says a user can add their organisation’s own SOPs or standards to what it checks against. It returns an executive summary followed by “hard violations and soft flags, each linked to the element with context on why it was flagged.” It is “rolling out gradually,” and the post tells users who cannot see it to contact their Planera team.
The premise is stated in a single line: “Passing every quality check doesn’t rule out finishes landing ahead of rough-ins.” That is a precise description of what schedule quality metrics do and do not measure. A DCMA-14-style check — Planera lists DCMA-14 quality checks among its homepage features — tests the mechanics of a network: missing logic, dangling activities, excessive lags, float distribution. It is indifferent to whether the sequence makes sense as a build. A plan can pass every one of those tests and still have a trade arriving before the work it depends on.
Separating hard violations from soft flags is the part that will determine whether the agent is used. A checker that returns a flat list of possible problems becomes noise on a large schedule; one that distinguishes a rule broken from a pattern worth a look is a tool a scheduler can run on every update.
The Procore Agent, and the Sentence That Matters
The third addition is headed “See which open RFIs could delay upcoming work.” It takes open Procore RFIs and, using Planera’s critical path and look-ahead, identifies which of them put upcoming work at risk. For each matched activity it writes a note carrying the RFI number, the responsible party, the due date and a link back to Procore. Where its confidence is low, the post says it reports the RFI as unmatched rather than guessing at a match.
Then this: “It only annotates, so dates and logic still go through Planera’s change controls.”
That is a deliberate ceiling, and it is placed at the exact point where the industry’s scheduling AI is currently splitting. Dated the same day, 9 November, Trimble announced Trimble Connect Scheduling at its Dimensions conference, one of ten AI capabilities it gave a stated availability status. Trimble’s release describes that scheduler as an “agentic planning engine” that links master CPM timelines to live jobsite data including field look-aheads and RFIs, and says it “continuously evaluates risk, recommends mitigation and puts new plans into action.” The same release describes its change-order automation as drafting change orders for human review and approval. The scheduling sentence carries no equivalent qualifier.
So two documents carrying the same date describe an RFI-to-critical-path capability and then describe its consequences in opposite terms. Trimble’s puts new plans into action. Planera’s only annotates. Both descriptions are the companies’ own, and this publication is reporting them rather than adjudicating between the products.
The reason this publication reads that fence as more consequential in scheduling than in document management or ERP is that a construction programme is a contractual artefact. Dates in an accepted baseline carry liquidated damages, extension-of-time claims and float ownership arguments behind them. An agent that revises a date has revised a position in a potential dispute. An agent that writes a note next to a date has not. Routing every change through “Planera’s change controls” is less a product limitation than a statement about who signs the programme.
Where Planera Is Coming From
Planera has raised $26.5 million in total, by its own account: a $5.4 million seed announced on 17 May 2023 from Sorenson Ventures and Firebolt Ventures, a $13.5 million Series A led by Sierra Ventures announced in August 2024 with Sorenson Capital, Brick and Mortar Ventures, Prudence VC and Firebolt Ventures participating, and $8 million more announced on 13 October 2025, with continued backing from those firms plus investment from Zachry Construction Corporation. The $26.5 million total is the figure that release gives. The company was founded by Nitin Bhandari, its chief executive, and Saif Lodhi, whom the seed release describes as a co-founder with more than 30 years in construction. The 2025 release was built around data centres: it named HITT Contracting, Ralph L. Wadsworth and Ryan Companies among its data centre customers, and it put the company’s growth in data centre projects at 600% over six months.
The relevance of that to this month’s release is that data centre work runs to compressed programmes with a great deal of design still moving — a combination that punishes sequencing errors and leaves a lot of open questions sitting on the critical path. A constructability checker and an RFI-to-float agent are both products you build for customers in that position.
What the Post Does Not Say
The post attaches no rollout language to the Procore agent or to the MCP — it marks the Constructability Agent as rolling out gradually and says nothing comparable about the other two, so their availability has to be inferred rather than read. It gives no pricing for any of the three, and no statement of whether the MCP connector is included in a subscription or sold separately. No accuracy figure appears for the RFI matcher, which matters given that it writes notes onto activities: the post says low-confidence RFIs are reported as unmatched, but does not say how often that happens or what share of matches are right.
Nor is the MCP’s scope enumerated the way ProjectTeam’s was. “Read-only and only sees what your Planera login can see” is a clear boundary, but it is a sentence rather than a list, and a buyer comparing connectors would want to know which schedule objects are exposed.
The three AI items also arrive alongside three more for Manny, Planera’s existing AI scheduling assistant, which can now take PDF uploads as chat context, render interactive HTML outputs inside Planera, and create code categories and values. The header image leads with the agents and the connector; the assistant is where the month’s other AI work went.
