Public API
26 min read
Every name you can import from shipit_agent, grouped by the module that defines it,
with the first line of its docstring. Generated by scripts/build_api_reference.py
from shipit_agent.__all__, so nothing exported is missing.
from shipit_agent import Agent, FunctionTool, PendingActionshipit_agent
| Name | Kind | What it is |
|---|---|---|
DEEP_AGENT_PROMPT | constant | A str. |
DEFAULT_AGENT_PROMPT | constant | A str. |
DEFAULT_SKILLS_PATH | constant | A Path. |
MCP_CATALOG | constant | A LazyCatalog. |
MODEL_ALIASES | constant | A dict. |
MODEL_PRICING | constant | A dict. |
STREAM_GENERATION | constant | A str. |
ToolRunOutput | type alias | A type alias. |
shipit_agent.agent
| Name | Kind | What it is |
|---|---|---|
Agent | class | The primary SHIPIT agent — LLM + tools + skills + RAG in one class. |
shipit_agent.agent_setup
| Name | Kind | What it is |
|---|---|---|
AgentDoctor | class | Inspect an agent's configuration and report what is missing or misconfigured. |
AgentProfile | class | A named agent configuration: prompt, tools and settings to build an agent from. |
AgentProfileBuilder | class | Build an :class:AgentProfile step by step with chained calls. |
DoctorCheck | class | The outcome of one health check: its name, status and a message. |
DoctorReport | class | Every check from one doctor run, with whether the agent passed overall. |
get_builtin_tool_map | function | Create all built-in tools and return them as a {name: tool} dict. |
get_builtin_tools | function | Convenience wrapper — returns built-in tools as a list. |
PromptTemplate | class | Renders prompt templates with {variable.path} references. |
shipit_agent.agents
| Name | Kind | What it is |
|---|---|---|
DelegationAdvice | class | Whether this task looks like sub-agent work, and why. |
DelegationPolicy | class | When to push the agent toward sub-agents, and how hard. |
shipit_agent.approvals
| Name | Kind | What it is |
|---|---|---|
ActionState | class | Where a queued action stands: pending, approved, rejected and so on. |
ApprovalQueue | class | Holds side-effecting calls until a human decides. |
PendingAction | class | One side-effecting call, held for a human decision. |
shipit_agent.batch
| Name | Kind | What it is |
|---|---|---|
BatchRequest | class | A single request to include in a message batch. |
BatchResult | class | The outcome of a single batched request. |
BatchRuntime | class | Runner for the Anthropic Messages Batches API. |
MessageBatchRunner | class | Runner for the Anthropic Messages Batches API. |
shipit_agent.chat_session
| Name | Kind | What it is |
|---|---|---|
AgentChatSession | class | AgentChatSession(agent: "'Agent'", session_id: 'str', trace_id: 'str | None' = None, session_store: 'SessionStore | None' = None, event_callbacks: 'list[EventCallback]' = |
shipit_agent.computer_use
| Name | Kind | What it is |
|---|---|---|
ActionKind | class | The kinds of actions a ComputerUseAgent can emit. |
ActionRecord | class | One action + the screenshot the agent saw before deciding it. |
BrowserSession | class | Minimum surface a browser must expose for the agent to drive it. |
ComputerAction | class | A structured action the agent decided to take. |
ComputerUseAgent | class | Drive a browser by showing screenshots to a vision-capable LLM. |
ComputerUseResult | class | Outcome of a full ComputerUseAgent.run(). |
MockBrowserSession | class | Deterministic test double — records every call and returns canned screenshots. |
parse_action | function | Best-effort parse of an LLM response into a ComputerAction. |
PlaywrightBrowserSession | class | Real-browser implementation backed by Playwright. |
shipit_agent.connections
| Name | Kind | What it is |
|---|---|---|
AuthKind | class | How a connection is authenticated — which decides how to fix it. |
Connection | class | One connectable resource and its current state. |
ConnectionRegistry | class | The connection state of everything this agent could reach. |
ConnectionState | class | Where a connection stands. |
shipit_agent.connectors
| Name | Kind | What it is |
|---|---|---|
connect | function | Turn a catalog connector into a live MCP server. |
connect_mcp | function | Turn a catalog connector into a live MCP server. |
Connector | class | A declarative integration definition. |
connector_categories | function | Every connector category in the catalog, sorted. |
list_connectors | function | Every connector, alphabetically; optionally filtered to one category. |
MCPCatalogEntry | class | A declarative integration definition. |
shipit_agent.context
| Name | Kind | What it is |
|---|---|---|
CompactionCheckpoint | class | Immutable record of one compaction. |
Compactor | class | Decides when to compact, and writes the handoff. |
ContextSnapshot | class | Point-in-time snapshot of context window utilization. |
ContextTracker | class | Tracks context window utilization across agent execution. |
count_message_tokens | function | Real token count across a message list, summing per-message content. |
count_tokens | function | Real token count of text for model, or the chars/4 estimate. |
get_model_limits | function | Limits for model, by longest-prefix match, with a safe default. |
TokenCalibrator | class | Per-model EMA of actual_tokens / estimated_tokens. |
shipit_agent.core
| Name | Kind | What it is |
|---|---|---|
AgentEvent | class | One observable moment of a run. |
AgentResult | class | Everything a finished run produced. |
Artifact | class | A file a tool left behind. |
AsyncAgentRuntime | class | Async version of AgentRuntime for use with asyncio/FastAPI/Starlette. |
build_schema_prompt | function | Build a prompt addition that instructs the LLM to output structured JSON. |
build_tool_schemas | function | The JSON schemas for every tool in registry, as sent to the model. |
construct_tool_registry | function | Build the registry of tools a run can call, including MCP tools. |
DuplicateToolError | class | Raised when two tools share the same public name. |
event_packet | function | A run event as a JSON-ready dict. |
is_pydantic_model | function | Check if obj is a Pydantic model class. |
Message | class | One turn of the conversation. |
parse_structured_output | function | Parse LLM output text according to the given schema. |
ReasoningResult | class | What a reasoning run produced: each reasoning tool's output and the events. |
ReasoningRuntime | class | Run an agent's reasoning tools (decompose, weigh options, synthesise) on one prompt. |
result_packet | function | A run result as a JSON-ready dict. |
RetryPolicy | class | Which failures are retried, how many times, and how long to wait. |
RouterPolicy | class | When a run plans first: automatic planning and the words that trigger it. |
schema_to_response_format | function | Convert an output_schema (Pydantic model or dict) to a response_format dict. |
ShipitAgentError | class | Base package exception. |
sse_event_packet | function | A run event as one Server-Sent Events frame. |
sse_result_packet | function | A run result as one Server-Sent Events frame. |
StructuredOutput | class | Wraps an LLM to produce reliable structured output. |
StructuredOutputResult | class | Result of a structured-output run. |
ToolCall | class | A model's request to run one tool. |
ToolRegistry | class | The tools a run can call, by name, including tools discovered from MCP servers. |
ToolResult | class | What a tool produced, bound to the call that asked for it. |
ToolRunner | class | Execute tool calls against a registry and collect their results. |
ToolRunnerResult | class | The results of a batch of tool calls, and any tools that were not found. |
websocket_event_packet | function | A run event shaped for a WebSocket message. |
websocket_result_packet | function | A run result shaped for a WebSocket message. |
shipit_agent.costs
| Name | Kind | What it is |
|---|---|---|
Budget | class | Cost budget for an agent run. |
BudgetExceededError | class | Raised when an agent exceeds its cost budget. |
CostRecord | class | Record of a single LLM call's cost. |
CostTracker | class | Real-time cost tracking with per-model pricing and budget enforcement. |
shipit_agent.deep
| Name | Kind | What it is |
|---|---|---|
AdaptiveAgent | class | Agent that can create new tools at runtime. |
AgentBenchmark | class | Benchmark framework for systematically testing agents. |
AgentMessage | class | Typed message for agent-to-agent communication. |
BenchmarkReport | class | Summary report of a benchmark run. |
Channel | class | Typed communication channel between agents. |
Checkpoint | class | A saved agent state. |
create_deep_agent | function | LangChain-style create_deep_agent factory function — but more powerful. |
DeepAgent | class | Power-user create_deep_agent-style factory. |
Goal | class | A structured goal with success criteria. |
GoalAgent | class | Autonomous agent that decomposes goals, tracks progress, and self-corrects. |
GoalResult | class | Result of a goal-directed agent run. |
PersistentAgent | class | Agent that can checkpoint and resume across sessions. |
ReflectionResult | class | Result of a reflective agent run. |
ReflectiveAgent | class | Agent that evaluates its own output and improves it through reflection. |
Supervisor | class | Hierarchical agent that plans, delegates, reviews, and combines results. |
SupervisorResult | class | Result of a supervisor run. |
TestCase | class | A single test case for agent benchmarking. |
Worker | class | A worker agent managed by a supervisor. |
shipit_agent.integrations
| Name | Kind | What it is |
|---|---|---|
CredentialRecord | class | One stored credential for a service: its secrets and any metadata. |
CredentialStore | class | Where connector tools look up credentials by service name. |
FileCredentialStore | class | File-backed credential store. |
FileOAuthStateStore | class | An OAuth state store kept in a JSON file, for flows that span processes. |
GoogleOAuthHelper | class | OAuth helper preconfigured for Google accounts. |
InMemoryCredentialStore | class | A credential store that lives for the life of the process. |
InMemoryOAuthStateStore | class | An OAuth state store that lives for the life of the process. |
OAuthClientConfig | class | An OAuth client's endpoints, client id and secret, scopes and redirect URI. |
OAuthHelper | class | Run the OAuth authorization-code flow for one provider. |
OAuthStateStore | class | Where pending OAuth state values are kept between redirect and callback. |
SlackOAuthHelper | class | OAuth helper preconfigured for Slack workspaces. |
shipit_agent.mcp
| Name | Kind | What it is |
|---|---|---|
discover_mcp_tools | function | The tools server offers. |
MCPError | class | An MCP server returned an error, timed out, or could not be reached. |
MCPHTTPTransport | class | Talk to an MCP server over HTTP with JSON-RPC POST requests. |
MCPPrompt | class | A prompt template exposed by an MCP server. |
MCPResource | class | A resource exposed by an MCP server (a file, table, doc, …). |
MCPServer | class | A named group of MCP tools that can be attached to an agent. |
MCPStdioTransport | class | Talk to an MCP server over stdio, starting a fresh process for each request. |
MCPStreamableHTTPTransport | class | Streamable-HTTP MCP transport (the 2025 spec revision). |
MCPSubprocessTransport | class | Talk to an MCP server over stdio, starting a fresh process for each request. |
MCPTool | class | A tool implemented in this process and served through an :class:MCPServer. |
PersistentMCPSession | class | Talk to an MCP server over stdio, keeping one process alive across requests. |
PersistentMCPSubprocessTransport | class | Talk to an MCP server over stdio, keeping one process alive across requests. |
RemoteMCPServer | class | An MCP server reached through a transport. |
shipit_agent.mcp_catalog
| Name | Kind | What it is |
|---|---|---|
list_mcp_catalog | function | All prebuilt connectors, alphabetically (optionally one category). |
shipit_agent.mcp_resilience
| Name | Kind | What it is |
|---|---|---|
CircuitBreaker | class | Per-server open/closed gate with exponential cooldown. |
ResilientMCPTransport | class | Wrap a transport with a circuit breaker + one bounded retry. |
shipit_agent.memory
| Name | Kind | What it is |
|---|---|---|
AgentMemory | class | Unified memory system combining conversation, semantic, and entity memory. |
ConsolidationResult | class | Return value of MemoryConsolidator.consolidate. |
ConversationMemory | class | Conversation memory with multiple strategies. |
DistilledFact | class | A single fact extracted by consolidation. |
Entity | class | A tracked entity (person, project, concept). |
EntityMemory | class | Track entities (people, projects, concepts) across conversations. |
InMemoryVectorStore | class | Simple in-memory vector store using cosine similarity. |
MemoryConsolidator | class | Run distillation, decay, and core-memory promotion on an AgentMemory. |
SearchResult | class | A single search result from semantic memory. |
SemanticMemory | class | Embedding-based memory for semantic search. |
VectorStore | class | Protocol for vector stores. |
shipit_agent.multimodal
| Name | Kind | What it is |
|---|---|---|
build_multimodal_message | function | Emit an Anthropic-shape message dict from a parsed prompt. |
extract_media_refs | function | Convenience: parse + return only the media references. |
FileMediaStore | class | JSON-backed store. Persists across process restarts. |
InMemoryMediaStore | class | Pure in-memory store. Fast, ephemeral, perfect for tests. |
MediaKind | class | Top-level kind of media a reference points to. |
MediaParser | class | Configurable parser. Construct once, reuse across messages. |
MediaReference | class | One media item extracted from a user prompt. |
MediaStore | class | Lookup interface for resolving media IDs to StoredMedia. |
ParsedPrompt | class | Result of parsing a user prompt for media references. |
StoredMedia | class | One uploaded asset, looked up by ID. |
shipit_agent.narrate
| Name | Kind | What it is |
|---|---|---|
build_transcript | function | A finished run's events → ordered transcript rows. |
NarratorRenderer | class | Render a live agent.stream(...) in the modern transcript style. |
render_transcript | function | Render a finished run as a plain-text transcript. |
render_transcript_html | function | Render a run's events as one self-contained HTML document. |
render_tree | function | Render a finished run as a tree, with no escape codes. |
StreamingToolInputParser | class | Incrementally parse a JSON object, streaming one string field. |
summarize | function | Narrate a single tool call. |
write_transcript | function | Write the transcript to path and return the resolved path. |
shipit_agent.notifications
| Name | Kind | What it is |
|---|---|---|
DiscordNotifier | class | Send notifications to Discord via a Webhook URL. |
Notification | class | A single notification to be sent to one or more channels. |
NotificationManager | class | Unified notification manager that dispatches to multiple channels. |
Notifier | class | Protocol that every notification channel must implement. |
SlackNotifier | class | Send notifications to Slack via an Incoming Webhook URL. |
TelegramNotifier | class | Send notifications to Telegram via the Bot API. |
shipit_agent.observability
| Name | Kind | What it is |
|---|---|---|
classify | function | How a client should treat this event across a reconnect. |
Durability | class | Whether a client should keep this after a reconnect. |
FileTraceStore | class | A trace store that writes one JSON file per trace in a folder. |
format_activity | function | Render a full run as a clean activity trace. |
format_event_line | function | Render one event as a display line (None = not user-facing). |
InMemoryTraceStore | class | A trace store that lives for the life of the process. |
sse | function | One Server-Sent Event. |
StreamFrame | class | One event, framed for the wire. |
StreamRenderer | class | Render a live event stream like Claude Code — tokens + tool cards. |
TraceRecord | class | Every event recorded for one trace, with its metadata. |
TraceStore | class | Where run events are recorded and read back by trace id. |
shipit_agent.parsers
| Name | Kind | What it is |
|---|---|---|
JSONParser | class | Parse JSON from LLM output. |
MarkdownParser | class | Parse structured content from markdown-formatted LLM output. |
MarkdownResult | class | Parsed sections from markdown text. |
OutputParser | class | Protocol for output parsers. |
parse_partial_json | function | Parse possibly-incomplete JSON. Returns the best-effort object, or None. |
ParseError | class | Raised when an output parser cannot parse the given text. |
PydanticParser | class | Parse LLM output into a Pydantic model instance. |
RegexParser | class | Parse LLM output using regex patterns. |
shipit_agent.pipeline
| Name | Kind | What it is |
|---|---|---|
parallel | function | Group steps to run concurrently. |
Pipeline | class | Deterministic composition of agents and functions. |
Step | class | A single step in a pipeline. |
step | function | Create a pipeline step. |
shipit_agent.rag
| Name | Kind | What it is |
|---|---|---|
CallableEmbedder | class | Wrap a user-supplied fn(list[str]) -> list[list[float]]. |
DocumentChunker | class | Greedy sentence-based chunker. |
HashingEmbedder | class | Deterministic feature-hashing embedder. |
HybridSearchPipeline | class | Orchestrates vector + keyword retrieval, RRF fusion, and post-processing. |
LLMReranker | class | Reranker that delegates relevance scoring to an LLM. |
RAG | class | The main RAG facade. |
RAGChunk | class | A single retrievable chunk of a document. |
RAGContext | class | A RAG search response. |
RAGDocument | class | A source document before chunking. |
RAGIndexFilters | class | Filters applied to vector/keyword search. |
RAGSearchQuery | class | A query submitted to the hybrid search pipeline. |
RAGSearchResult | class | A ranked chunk returned by the search pipeline. |
RAGSource | class | A source citation attached to an AgentResult after a RAG-backed run. |
shipit_agent.replay
| Name | Kind | What it is |
|---|---|---|
diff_traces | function | Diff two traces. Accepts either TraceReplayer or raw TraceRecord. |
ForkPoint | class | Describes how a trace was forked. |
ReplayCheckpoint | class | State captured at a fork point — enough to resume an agent run. |
ReplayResult | class | The outcome of resuming a fork — wraps an AgentResult. |
TraceDiff | class | Result of comparing two traces. |
TraceReplayer | class | Wraps a TraceRecord to enable inspection + forked replay. |
shipit_agent.safety
| Name | Kind | What it is |
|---|---|---|
AgentHooks | class | Lifecycle hooks for an agent — observe, and now block or rewrite. |
console_permission_prompt | function | Build a permission_callback that asks the human on the console. |
format_call_preview | function | A one-line preview of a tool call for an approval prompt. |
GuardDecision | class | Outcome of a guardrail check. |
Guardrails | class | Composable content guardrails for prompts, outputs, and tool args. |
LockdownPolicy | class | When to latch, and what latching means. |
LockdownState | class | Whether this run has latched, and why. |
PermissionDecision | class | What a permission rule decides about a tool call: allow, deny or ask. |
PermissionEngine | class | Rule-based gate for tool calls. |
PermissionResult | class | The outcome of a permission check. |
shipit_agent.scheduling
| Name | Kind | What it is |
|---|---|---|
AgentScheduler | class | Run an agent on a recurring schedule — cron jobs for agents. |
fire_all | function | Queue a batch — one poll of an inbox, one page of rows. |
InMemoryTriggerQueue | class | For tests and single-process toys. |
ScheduledAgentConfig | class | Serializable agent/runtime selection for one scheduled job. |
ScheduledAgentFactory | class | Build isolated agents from persisted scheduled-job configuration. |
ScheduledJob | class | A recurring agent job — a prompt plus when to run it. |
ScheduleResult | class | Result envelope for a scheduled agent execution. |
ScheduleRunner | class | Execute an agent or session-backed agent call on behalf of a scheduler. |
SQLiteJobStore | class | Persist scheduler jobs so they survive process restarts. |
SqliteTriggerQueue | class | The default: events survive the process that received them. |
Trigger | class | A source, and what to ask the agent when it fires. |
TriggerEvent | class | One thing that happened, waiting to be turned into a run. |
TriggerRegistry | class | What is wired to what, and the loop that delivers it. |
TriggerRun | class | What one delivery did. |
shipit_agent.sessions
| Name | Kind | What it is |
|---|---|---|
SessionManager | class | Manage chat session lifecycle on top of a SessionStore. |
shipit_agent.skills
| Name | Kind | What it is |
|---|---|---|
apply_skill | function | Merge a skill's prompt text into the agent's system prompt. |
create_skill | function | Create a Skill with practical defaults for local authoring. |
discover_project_skills | function | Discover local SKILL.md files in common agent skill directories. |
FileSkillRegistry | class | A :class:SkillRegistry backed by a JSON file. |
find_relevant_skills | function | Return the most relevant skills for a user message. |
load_markdown_skill | function | Load a Codex/Claude-style SKILL.md into a runtime Skill. |
match_skill_by_trigger | function | Find the best skill whose trigger phrases appear in the user message. |
Skill | class | A single agent skill with metadata, configuration, and marketplace info. |
skill_id_from_name | function | Create a stable slug-style skill id from a display name. |
SkillCatalog | class | Create and persist skills in a local JSON catalog. |
SkillRegistry | class | In-memory registry of :class:Skill objects. |
shipit_agent.stores
| Name | Kind | What it is |
|---|---|---|
FileMemoryStore | class | A memory store kept in one JSON file. |
FileSessionStore | class | A session store that writes one JSON file per session in a folder. |
InMemoryMemoryStore | class | A memory store that lives for the life of the process. |
InMemorySessionStore | class | A session store that lives for the life of the process. |
MemoryFact | class | One remembered fact, with its category and metadata. |
MemoryStore | class | Where an agent saves facts and searches them later. |
SessionRecord | class | A saved conversation: its messages and metadata. |
SessionStore | class | Where an agent loads and saves conversations by session id. |
shipit_agent.team
| Name | Kind | What it is |
|---|---|---|
AgentTeam | class | Multi-agent team with LLM-routed coordination. |
TeamAgent | class | An agent with a role in a team. |
TeamResult | class | Result of a team run. |
TeamRound | class | Record of one round of team coordination. |
shipit_agent.tools
| Name | Kind | What it is |
|---|---|---|
AgentTool | class | Expose an Agent-compatible object as one focused callable tool. |
ArtifactBuilderTool | class | The build_artifact tool: a deliverable with a stable id, versions and in-place updates. |
AskUserTool | class | The ask_user tool: request structured user input as part of a human-in-the-loop flow. |
BashTool | class | The bash tool: run any shell command in the project root. |
BraveSearchProvider | class | Web search through the Brave Search API. |
build_search_provider | function | The search provider named by provider (or the one passed in), configured with api_key. |
ClaudeMemoryTool | class | Anthropic-style memory_20250818 tool. |
CodeExecutionTool | class | The run_code tool: execute Python or shell code in a local subprocess workspace. |
ConfluenceTool | class | The confluence tool: search and create Confluence pages. |
contract_for | function | The contract for a tool: declaration → registration → table → guess. |
CustomAPITool | class | The custom_api tool: call a configured custom HTTP API. |
DecisionMatrixTool | class | The decision_matrix tool: compare options against explicit criteria and recommend the strongest choice. |
DeepResearchTool | class | The deep_research tool: deep multi-angle research: several search queries, deduped sources, top pages fetched — returns a structured digest with citations. |
DocumentBuilderTool | class | The build_document tool: create a polished PDF report, Excel workbook, Word document, PowerPoint deck, or styled HTML page from structured content. |
DownloadFileTool | class | The download_file tool: download a file (binary-safe: zip, csv, image, pdf, …) from an http(s) URL to local disk. |
DuckDuckGoSearchProvider | class | Web search through DuckDuckGo's HTML results; needs no API key. |
EditFileTool | class | The edit_file tool: apply an exact string replacement patch to an existing file. |
EvidenceSynthesisTool | class | The synthesize_evidence tool: turn raw observations into facts, inferences, gaps, and recommendations. |
FigmaTool | class | The figma tool: read Figma files, render node images, browse and post comments, and introspect team projects / components via the Figma REST API. |
FileReadTool | class | The read_file tool: read a file from the local project. |
FileWriteTool | class | The write_file tool: create or overwrite a file in the local project. |
FunctionTool | class | A tool that calls a Python function. |
GitHubTool | class | The github tool: work with GitHub issues, pull requests, file contents, and Actions workflow runs via the REST API. |
GitLabTool | class | The gitlab tool: search GitLab issues, manage merge requests, read files, and control CI pipelines. |
GitOpsTool | class | The git_ops tool: structured git operations: status, diff, log, show, blame, branch, add, commit, checkout, stash (push/reset are disabled by default). |
GlobSearchTool | class | The glob_files tool: find files in the local project by glob pattern. |
GmailTool | class | The gmail_search tool: search, read, draft, and send Gmail messages from a configured Google account. |
GoogleCalendarTool | class | The google_calendar tool: search and list Google Calendar events. |
GoogleDriveTool | class | The google_drive tool: search and list Google Drive files. |
GoogleSheetsTool | class | The google_sheets tool: read / write Google Sheets cells, ranges, formulas, and sheet structure. |
GrepSearchTool | class | The grep_files tool: search file contents in the local project using ripgrep when available. |
HumanReviewTool | class | The human_review tool: pause for human review and approval before continuing. |
JiraTool | class | The jira tool: search Jira issues and create Jira issues. |
LinearTool | class | The linear tool: search Linear issues and create Linear issues. |
LinkedInSearchTool | class | Read-only LinkedIn lookup. |
MemoryTool | class | The memory tool: store and retrieve structured memory facts for the current agent. |
NotebookEditTool | class | The notebook_edit tool: read, edit, add, or delete cells in a Jupyter notebook (.ipynb). |
NotionTool | class | The notion tool: search Notion pages and create Notion pages. |
OpenURLTool | class | Fetch a URL and return clean text. |
PDFTool | class | The pdf tool: extract text, per-page content, or metadata from a PDF file (local or URL). |
PlannerTool | class | The plan_task tool: generate a concrete execution plan with ordered steps, risks, and checkpoints. |
PlaywrightBrowserTool | class | The playwright_browse tool: use Playwright to open a page and return the rendered text content. |
PlaywrightSearchProvider | class | Web search by driving a real browser with Playwright. |
PromptTool | class | The build_prompt tool: build or refine a system prompt from goals, constraints, and style instructions. |
SalesforceTool | class | The salesforce tool: search Salesforce records; query SOQL/SOSL; read accounts/opportunities/contacts; log activities. |
SerperSearchProvider | class | Web search through the Serper (Google results) API. |
SlackTool | class | The slack tool: search Slack messages and post Slack messages. |
SQLTool | class | Single SQL tool for every SQLAlchemy dialect. |
StripeTool | class | The stripe tool: read-heavy Stripe data: customers, charges, subscriptions, invoices, products. |
SubAgentTool | class | Delegate a focused task to a real agent — foreground or background. |
TavilySearchProvider | class | Web search through the Tavily API, built for agents. |
ThoughtDecompositionTool | class | The decompose_problem tool: break a problem into workstreams, assumptions, risks, evidence needs, and next actions. |
TodoTool | class | Live task checklist — the SHIPIT equivalent of Claude Code's TodoWrite. |
Tool | class | The protocol every tool implements. |
ToolActionKind | class | A stable machine tag paired with the label a human reads. |
ToolContext | class | What a tool can see while it runs. |
ToolContract | class | What a tool is, so nothing downstream has to infer it. |
ToolOutput | class | A complete tool result plus an optional compact model-facing view. |
ToolOutputChunk | class | One incremental piece of a streaming tool result. |
ToolSearchTool | class | Lexical tool discovery for agents with many available tools. |
VerifierTool | class | The verify_output tool: check whether content satisfies a set of required criteria. |
VisionTool | class | Send an image + question to a vision-capable LLM, return the analysis. |
WebhookPayloadTool | class | Expose the triggering webhook payload to the agent. |
WebSearchTool | class | The web_search tool: search the web and return structured search results. |
WorkspaceFilesTool | class | The workspace_files tool: read, write, list, and inspect files in the local shipit workspace. |
ZendeskTool | class | The zendesk tool: search / read / comment on Zendesk tickets and users; list and preview macros. |
shipit_agent.verifier
| Name | Kind | What it is |
|---|---|---|
PreToolDecision | class | Verifier's decision about a single proposed tool call. |
PreToolVerifier | class | Calls a verifier LLM before each tool execution. |
ProgressCheck | class | Verifier's rating of one agent iteration. |
ProgressVerifier | class | Per-iteration "did we make progress?" check. |
VerifierConfig | class | Options for VerifierNetwork. Passed when constructing one of the verifier classes. |
VerifierNetwork | class | Combined pre-tool + progress verifier for an agent run. |
VerifierVerdict | class | Verdict returned by the pre-tool verifier for a proposed tool call. |
shipit_agent.workspace
| Name | Kind | What it is |
|---|---|---|
discover_commands | function | Map command name -> .md path under .shipit/commands/. |
expand_command | function | Expand a /command invocation into its prompt, or return None. |
load_project_memory | function | Return a single prompt block of project + user instruction files. |
load_settings | function | Merge ~/.shipit/settings.json (user) then .shipit/settings.json. |
WorkspaceSettings | class | Per-project agent settings: permission mode and allow/deny rules. |