By David Speakman ·
SPEAKMAN.AI is the free, local, open source engine behind the same MCP workflows the hosted platform runs. No account and no cloud dependency beyond your own API key. Version 1.1.0 adds a middle model tier, native Vertex AI configuration, and a new workflow that turns a business description into implementation ready technical contracts.
Every agent in every workflow now resolves against three positions instead of two, and providers or models are swapped centrally with no workflow files to edit.
Fast or advanced meant every mid-weight task got routed to whichever end of the range was closest, and simple work got overpaid for while harder work went underpowered. A standard tier sits between them, and every provider now exposes all three positions consistently, including the ones that months of internal use kept surfacing a need for: Vertex AI, and the newest Gemini and Claude model families.
This release also reflects that internal use directly. A workflow seeding rule that only ever
ran once, silently skipping any bundled workflow added after your first launch. A model tier
resolver that quietly ignored a configured fast model for every provider except Ollama, plus
a stale fallback default left over from before the tier split existed. A desktop build that
could inherit Docker-only credential settings from an unrelated .env file. None
of these show up in a demo, they show up when you run the thing for months.
A configurable middle position, STANDARD_MODEL, between fast and advanced, resolved per provider with no workflow files to edit.
Vertex AI is now configured right in the setup page: GCP project and region, one click, authenticating via Application Default Credentials so there's no API key to type or leak.
A new built-in pipeline. A business description goes in; a domain model, validated use cases, field definitions, and a full OpenAPI schema come out. Detailed below.
New bundled workflows are picked up on your next launch automatically, with no database reset and nothing you've customized touched in the process.
Point any tier at the Gemini 3 or Claude 5 model families when the extra quality is worth the extra cost. Still opt in, never the default.
A collapsed Anthropic tier map, an ignored fast-model override, a stale fallback default from before the tier split, and a Windows-only credentials bug, all found by running the app rather than reading the code.
Seeded from a completed Solution Architecture Document, the same JSON output
/generate-sad produces from a project's concepts and organizational
capabilities, MCP_REQUIREMENTS_ENGINEERING_V1 runs five further stages with a
validation loop at every handoff. Each stage is checked against the one before it for
coverage, not just presence.
This is the same workflow the /generate-requirements skill calls into as Phase 2
of the SPEAKMAN.AI SDLC pipeline, shown below.
The hosted SPEAKMAN.AI platform runs a full idea-to-production SDLC pipeline. This repository is the free, local, MIT licensed engine that executes the same MCP workflows, with full functionality for an individual user and no account required.
Business description in, Solution Architecture Document out.
Domain model, naming dictionary, use cases, API and DB schema.
A working, milestone gated, git committed codebase.
Terraform for GCP, AWS, and Azure, plus pre-pentest hardening.
A gated security engagement, run against the live staging build.
Google promoted a Gemini 3 preview model to general availability mid cycle and retired its preview alias. We hit the resulting 404 while testing this exact release, then traced it to the renamed model and shipped the fix.
gemini-3.1-flash-lite-preview → 404 Not Found
gemini-3.1-flash-lite → 200 OK
| Mode | Command | Database |
|---|---|---|
| Windows exe | Run SpeakmanAI.exe | ~/.speakmanai/speakmanai.db |
| Docker + SQLite | docker-compose up --build | speakmanai_data volume |
| Docker + MongoDB | docker-compose --profile mongo up | MongoDB, port 27017 |
| Dev, no Docker | uvicorn server:app --port 8000 | ~/.speakmanai/speakmanai.db |
There's no account, no cloud dependency, and no per-call fee beyond your own provider key. Workflows are plain JSON on disk, readable and editable rather than locked inside a hosted black box.
Download the exe, or clone and run with Docker. You get full functionality on your own API keys, with nothing sent anywhere else.
The hosted SPEAKMAN.AI platform runs the same engine with the infrastructure teams need on top of it.
Written by David Speakman. Speakman Consulting designs and builds this kind of system for growing organizations: agent workflows with the governance that keeps humans in the loop.