PromptisePromptise
Docs
GitHub
← Promptise> how it works

How Promptise handles one request, module by module.

A customer writes in at three in the morning. Follow the ticket through all twelve modules: what wakes the agent, what it knows, how it reasons, what it calls, who signs off and what it records. Every part links to the docs that explain it.

Step 01 / 1403:02:11 · Agent Runtime

A ticket wakes the agent. A helpdesk webhook fires. The runtime wakes the support agent's process and hands it the ticket.

An illustrative run. Names, amounts and timings are examples; what each step describes is what the framework does.

Reply: The answer, streamed as it is written.Record: What happened, kept: traces, the journal and the audit log.Actions: Changes made in your systems, gated where it matters.readsgeneratescallsrequestsblocksrecallsplansresultscheckspasseswakesacts asscansspendstool callruns codeOpenAPI spec: An API description you already have: the input to MCPcast.OPENAPI SPECMCPcast: Turns the spec into a reviewed, generated MCP server.MCPCASTPeer agents: Other agents this one can delegate to.PEER AGENTSMCP client: Discovers and calls tools on every connected server.MCP CLIENTSandbox: Where code the agent writes runs, isolated from your systems.SANDBOXOrders server: A server your team built with the MCP Server SDK.ORDERS SERVERBilling server: A server MCPcast generated from the billing API's OpenAPI document.BILLING SERVERYour APIs: The systems behind the servers.YOUR APISPrompt: Typed blocks with priorities: identity, rules, format, examples.PROMPTMemory: Relevant memories, searched and injected before the call.MEMORYContext: Everything the model will see, assembled inside a token budget.CONTEXTCache: A semantic cache of past answers.CACHEPlan: Breaks the task into subgoals.PLANAct: A model call with tools: where the agent reaches outward.ACTObserve: Interprets tool results and extracts the facts.OBSERVEValidate: A quality gate that can send the graph back a step.VALIDATESynthesize: Combines what was found into the final answer.SYNTHESIZETriggers: Cron, webhooks, file watches, events and messages.TRIGGERSProcess: The long-running process the agent lives in.PROCESSIdentity: Who the agent is, provable to the systems it calls.IDENTITYGuardrails: Scans what goes in and what comes out.GUARDRAILSBudget: How much the agent may do on its own.BUDGETHealth: Catches stuck agents, loops and error spikes.HEALTHMission: The goal, and whether the agent is getting there.MISSIONJournal: A durable record that makes crash recovery possible.JOURNALTraces: Every model turn and tool call, exported where you look.TRACES
InterfaceIt acts
AgentIt thinks
HarnessIt operates
Replystreamed to the userRecordtraces · journal · auditActionsapproved · executed
01 · Agentstandby
The Promptise Agentdocs ↗
  • build_agent()
  • Models & providers
  • Server specs
  • Memory providers
  • RAG foundation
  • Conversations
  • Semantic cache
  • Tool optimization
  • Model fallback
  • Adaptive strategy
  • Streaming
  • Events & notifications
  • Cross-agent delegation
  • SuperAgent files
  • Default system prompt
  • Callback handler
  • Environment resolver
  • Agent CLI
02 · Agentstandby
Reasoning Enginedocs ↗
  • Reasoning graph
  • Reasoning nodes
  • Standard nodes
  • Custom nodes
  • Node flags
  • Edges & transitions
  • Prebuilt patterns
  • Skills library
  • Skill registry
  • Patterns in build_agent
  • Custom reasoning guide
03 · Agentstandby
Execution Enginedocs ↗
  • Autonomous mode
  • Static mode
  • Execution loop
  • Runtime tool injection
  • Processors
  • Engine hooks
  • Graph mutation
  • Serialization & YAML
  • Code-action
04 · Agentstandby
Prompt & Contextdocs ↗
  • PromptBlocks
  • ConversationFlow
  • PromptBuilder
  • Strategies
  • Perspectives
  • Chaining
  • Context providers
  • Prompt guards
  • Inspector
  • Prompt testing
  • Suite & registry
  • Loader & templates
  • Shell context injection
  • Context engine
  • Context lifecycle
05 · Interfacestandby
MCP Server SDKdocs ↗
  • Server fundamentals
  • Routers
  • Middleware
  • Authentication
  • Role & scope guards
  • Dependency injection
  • Multi-tenancy
  • Approval gates
  • Result caching
  • Rate & concurrency limits
  • Resilience
  • Audit log
  • Metrics & tracing
  • Job queue
  • Versioning & composition
  • Deployment
  • TestClient
06 · Interfacestandby
MCP Clientdocs ↗
  • MCPClient
  • MCPMultiClient
  • Tool adapter
  • Tools & schemas
07 · Interfacestandby
MCPcastdocs ↗
  • Parse
  • Classify
  • Curate
  • Review
  • Emit
  • Measure
  • Safety profiles
  • Auth modes
  • Guided setup
08 · Harnessactive
Agent Runtimedocs ↗
  • Agent processesrunning
  • Runtime managerrunning
  • Lifecyclerunning
  • Cron trigger
  • Webhook triggerrunning
  • Event & message triggers
  • File-watch trigger
  • Context & state
  • Conversation buffer
  • Journal
  • Replay engine
  • Rewind engine
  • Runtime hooks
  • Open mode
  • Manifests
  • Configuration
  • Orchestration API
  • Message inbox
  • Distributed coordinator
  • Discovery & transport
  • Dashboard
  • Runtime CLI
09 · Harnessstandby
Governance & Controlsdocs ↗
  • Autonomy budget
  • Behavioral health
  • Missions
  • Secret scoping
  • Approval policy
  • Approval handlers
  • Auto-approval classifier
  • Irreversible actions
10 · Harnessstandby
Agent Identitydocs ↗
  • Local identity
  • Verifiable identity
  • Microsoft Entra ID
  • AWS IAM
  • Google Cloud
  • SPIFFE / SPIRE
  • Generic OIDC
  • Presented to MCP servers
  • CallerContext
  • Guarantees & threat model
11 · Harnessstandby
Guardrails & Sandboxdocs ↗
  • Injection detector
  • PII detector
  • Credential detector
  • Names & addresses
  • Content safety
  • Custom rules
  • Sandbox
  • Security profiles
  • Network modes
  • Memory sanitization
12 · Harnessstandby
Observabilitydocs ↗
  • Observe levels
  • Exporters
  • Trace capture
  • Post-run analysis
  • Multi-user attribution
  • Execution report
Ticket №4411webhook

“I was charged twice for order 4411. Can you refund one?”

froma.keller@example.com
scanpending

How to read it

  • A component. Click anything for its docs.
  • Where a module sits in the stack.
  • running: doing the work in this step.
  • done: used earlier in this run.
Ticket №4411webhook

“I was charged twice for order 4411. Can you refund one?”

froma.keller@example.com
scanpending
InterfaceIt acts
05 · 17 partsMCP Server SDKstandby
Module overview ↗
  • Server fundamentals
  • Routers
  • Middleware
  • Authentication
  • Role & scope guards
  • Dependency injection
  • Multi-tenancy
  • Approval gates
  • Result caching
  • Rate & concurrency limits
  • Resilience
  • Audit log
  • Metrics & tracing
  • Job queue
  • Versioning & composition
  • Deployment
  • TestClient
06 · 4 partsMCP Clientstandby
Module overview ↗
  • MCPClient
  • MCPMultiClient
  • Tool adapter
  • Tools & schemas
07 · 9 partsMCPcaststandby
Module overview ↗
  • Parse
  • Classify
  • Curate
  • Review
  • Emit
  • Measure
  • Safety profiles
  • Auth modes
  • Guided setup
AgentIt thinks
01 · 18 partsThe Promptise Agentstandby
Module overview ↗
  • build_agent()
  • Models & providers
  • Server specs
  • Memory providers
  • RAG foundation
  • Conversations
  • Semantic cache
  • Tool optimization
  • Model fallback
  • Adaptive strategy
  • Streaming
  • Events & notifications
  • Cross-agent delegation
  • SuperAgent files
  • Default system prompt
  • Callback handler
  • Environment resolver
  • Agent CLI
02 · 11 partsReasoning Enginestandby
Module overview ↗
  • Reasoning graph
  • Reasoning nodes
  • Standard nodes
  • Custom nodes
  • Node flags
  • Edges & transitions
  • Prebuilt patterns
  • Skills library
  • Skill registry
  • Patterns in build_agent
  • Custom reasoning guide
03 · 9 partsExecution Enginestandby
Module overview ↗
  • Autonomous mode
  • Static mode
  • Execution loop
  • Runtime tool injection
  • Processors
  • Engine hooks
  • Graph mutation
  • Serialization & YAML
  • Code-action
04 · 15 partsPrompt & Contextstandby
Module overview ↗
  • PromptBlocks
  • ConversationFlow
  • PromptBuilder
  • Strategies
  • Perspectives
  • Chaining
  • Context providers
  • Prompt guards
  • Inspector
  • Prompt testing
  • Suite & registry
  • Loader & templates
  • Shell context injection
  • Context engine
  • Context lifecycle
HarnessIt operates
08 · 22 partsAgent Runtimeactive
Module overview ↗
  • Agent processesrunning
  • Runtime managerrunning
  • Lifecyclerunning
  • Cron trigger
  • Webhook triggerrunning
  • Event & message triggers
  • File-watch trigger
  • Context & state
  • Conversation buffer
  • Journal
  • Replay engine
  • Rewind engine
  • Runtime hooks
  • Open mode
  • Manifests
  • Configuration
  • Orchestration API
  • Message inbox
  • Distributed coordinator
  • Discovery & transport
  • Dashboard
  • Runtime CLI
09 · 8 partsGovernance & Controlsstandby
Module overview ↗
  • Autonomy budget
  • Behavioral health
  • Missions
  • Secret scoping
  • Approval policy
  • Approval handlers
  • Auto-approval classifier
  • Irreversible actions
10 · 10 partsAgent Identitystandby
Module overview ↗
  • Local identity
  • Verifiable identity
  • Microsoft Entra ID
  • AWS IAM
  • Google Cloud
  • SPIFFE / SPIRE
  • Generic OIDC
  • Presented to MCP servers
  • CallerContext
  • Guarantees & threat model
11 · 10 partsGuardrails & Sandboxstandby
Module overview ↗
  • Injection detector
  • PII detector
  • Credential detector
  • Names & addresses
  • Content safety
  • Custom rules
  • Sandbox
  • Security profiles
  • Network modes
  • Memory sanitization
12 · 6 partsObservabilitystandby
Module overview ↗
  • Observe levels
  • Exporters
  • Trace capture
  • Post-run analysis
  • Multi-user attribution
  • Execution report

Get your first agent into production.

Open source and free under Apache 2.0, with no account to create. Install it with pip install promptise and follow the quickstart.

Read the quickstartGitHub
Promptise - AI Framework LogoPromptise

The foundation layer for agentic intelligence. Build, secure and operate autonomous AI systems with Promptise Foundry.

pip install promptise

[01] Foundry

  • MCPcast
  • The Promptise Agent
  • Reasoning Engine
  • MCP
  • Agent Runtime
  • Prompt Engineering
  • Execution Engine
  • Agent Identity

[02] Resources

  • Documentation
  • GitHub
  • Guides
  • Learning Paths
  • Questions

[03] Company

  • About
  • Terms of Service
  • Privacy Policy
  • Cookie Policy
  • Subprocessors

© 2026 Promptise by Manser Ventures. All rights reserved.

Open source · Python