Build AI agents that think and act on their own with just one install.
Joined October 2026
- Tweets39
- Following4
- Followers29
- Likes2
Pinned Tweet
$PROMPTISE is now live on @ponsdotfamily
CA: 0xc888e05fb384e9070217086f000d318a1533657d
Most agent tokens ship a chatbot and a whitepaper.
Promptise ships software: 13 releases, 12 modules, open source.
Guardrails, budgets and approval gates built in.
Questions teams ask before they adopt it.
Short answers. The docs have the long ones.
All six, one by one:
2/ Which models does it work with?
Any of them. Pass a string like "openai:gpt-5-mini" or "anthropic:claude-sonnet-4.5", or hand it any LangChain chat model. Configuration is in code, not a wall of environment variables.
1/ What exactly is in the package?
The agent framework, a production SDK for building MCP servers, a prompt-engineering layer and an agent runtime. MCPcast, which turns an existing OpenAPI document into a reviewed MCP server, ships with them.
Make your API usable by every AI assistant, safely.
Your customers are starting to work through AI assistants. MCPcast turns the OpenAPI document you already have into a reviewed MCP server in one command.
How it works, in eight parts:
8/ Get started
From OpenAPI to an agent using your API.
pip install promptise, then run promptise mcpcast. Point it at a spec, read what it decided, fix what it got wrong, and hand the result to Claude Desktop. The plan stays yours.
promptise.com/mcpcast
7/ Measure and ship
--eval writes tasks from your plan, runs a real agent against the full server pipeline and grades the run A to F: the right tool first, valid arguments, tools no task could reach, examples the API does not recognise.
6/ Authentication
Control whose credentials reach your API. Four deployments, one flag:
· --auth env-token: Personal — a desktop client
· --auth passthrough: Shared — behind your gateway
· --auth api-key: Multi-tenant product
· --auth none: Local demo
5/ Approvals
The gate is middleware in the server’s own request pipeline, so it applies to every client.
Fail closed. A client that cannot ask, a reviewer who declines, a decision that never comes: all end the same way with APPROVAL_DENIED, and nothing is sent upstream.
4/ Safety
What an agent may do is your decision, not an accident.
The safety profile decides which risk classes become tools at all. Everything that is not a read waits for a human, enforced inside the server for every MCP client.
3/ Review
A model designed this surface, so it will misread something. The warnings are computed, not guessed:
· a description that names a tool this plan does not expose
· parameters sent as fixed defaults
· examples that teach the wrong identifiers
2/ The pipeline
Automation where it is safe, your judgement where it counts.
Code does the deterministic work. A model designs the surface under rules code enforces. You review the plan before a single file is written.
Parse · Classify · Curate · Review · Emit · Measure
2/ What it is
An OpenAPI document describes everything your API can do: hundreds of operations, developer prose, parameters nobody sets.
Hand that to a model and it picks the wrong call, invents identifiers, and deletes something on a Tuesday.
From pip install to a running agent, in three steps:
- Install: one package. Add [all] for vector memory, ML guardrails and exporters.
- Build: point it at your MCP servers, pick a model.
- Run: every call is checked by guardrails and traced.
docs.promptise.com/getting-s…
Not a whitepaper. 12 modules you can pip install today:
Agent: reasoning engine, prompts as code
Interface: MCP server SDK, MCP client, MCPcast
Harness: runtime, governance, identity, guardrails + sandbox, observability
Open source, Apache 2.0.
promptise.com
Where $PROMPTISE is going, from the roadmap:
Shipped: 13 releases since 1.0. Latest is v1.10.1.
Now: new MCP protocol for agents + runtime
Next: evals, A2A protocol, dev UI, signed approvals
Later: one-command deploy, MCP hosting, control center
promptise.com/roadmap
Twelve modules. Start with the one you need.
Each one solves one production problem. Add the others when you need them: they share identity, audit and configuration, so nothing has to be glued together.
All twelve, one by one:
Start with one module. Add the rest when you need them.
Open source, one pip install, and nothing to glue together.
promptise.com/#modules
