# Felspar Docs

> Felspar turns your company's policy documents into tested, approved Decision Packs that every agent can share.

Felspar is a shared decision service for AI agents. A company's policy documents become tested, approved, versioned Decision Packs. Agents call list_jobs to pick a job, get_job_requirements and get_job_decision (MCP at https://mcp.felspar.ai/mcp, or REST at https://api.felspar.ai) with the case's ID and exact source excerpts, and receive the company's outcome, why, the next step, any typed escalation and a verifiable receipt; check_output checks text an agent wrote for a decision. Code that acts verifies the receipt with POST /v0/executions/{id}/verify first. Authoring tools for people: list_sources, read_source, add_current_version, get_pack_guide, check_pack, propose_pack, backtest. Approval and release happen only on the web app at https://app.felspar.ai.

## Get started

- [Introduction](https://docs.felspar.ai/): Felspar turns your company's policy documents into tested, approved decisions that every agent can share.
- [How it works](https://docs.felspar.ai/how-it-works): The full Felspar flow, from a policy document to a verified decision, in one diagram.
- [Quickstart](https://docs.felspar.ai/quickstart): From a Google Doc to a released Decision Pack and your first decision, in about 15 minutes.

## Core concepts

- [Decision Packs](https://docs.felspar.ai/concepts/decision-packs): What a Decision Pack is and the parts it is made of - job, purpose, subject, questions, answers, rules, outcomes, next steps, escalations and test cases.
- [Evidence and classification](https://docs.felspar.ai/concepts/evidence): How agents send evidence as exact excerpts, how Felspar reads numbers, dates and text, and what "needs evidence" means.
- [Proposals, approvals and releases](https://docs.felspar.ai/concepts/releases): How a pack goes from a draft to something agents follow, and how Latest and pinned releases work.
- [Receipts and verification](https://docs.felspar.ai/concepts/receipts): What a Felspar receipt records, how verification replays a decision, and what a receipt does and doesn't prove.
- [Subjects and escalations](https://docs.felspar.ai/concepts/subjects-and-escalations): What each case is about, which records relate to it, and the typed hand-offs a pack names for cases it can't decide.
- [The eleven categories](https://docs.felspar.ai/concepts/categories): The eleven categories every policy maps into, including subjects and relationships, and how a real policy passage lands in each.
- [Worked example - Fernwood](https://docs.felspar.ai/concepts/fernwood-example): A complete fictional policy, the pack made from it, and decisions on three companies, end to end.

## Guides

- [Connect Google Docs](https://docs.felspar.ai/guides/connect-google-docs): Connect your own Google Drive, add the Docs that define your decisions, and understand per-person access.
- [Draft a pack](https://docs.felspar.ai/guides/draft-a-pack): Turn a connected policy document into a proposed Decision Pack with ChatGPT or Claude and the Felspar plugin.
- [Write test cases](https://docs.felspar.ai/guides/write-test-cases): Write test cases that make a pack trustworthy - one per outcome, both sides of every boundary, every exception and missing evidence.
- [Review and backtest a change](https://docs.felspar.ai/guides/review-and-backtest): See what a proposed pack change would do to real recent decisions, and which agents it would reach, before anyone approves it.
- [Approve and release](https://docs.felspar.ai/guides/approve-and-release): Review a proposal on the Felspar web app, approve that exact proposal, and release it to your agents.
- [Source changes and triage](https://docs.felspar.ai/guides/source-changes): How Felspar watches your policy documents, what happens when one is edited, and how to triage the change.
- [Register an agent](https://docs.felspar.ai/guides/register-an-agent): Give a company agent its own identity and key, choose its jobs, and hand its builder the setup.
- [Retries and attempts](https://docs.felspar.ai/guides/retries-and-attempts): What an agent does when Felspar can't decide yet - the retry guidance, the attempt limit, and when to escalate.
- [The pre-action gate](https://docs.felspar.ai/guides/pre-action-gate): Check in code, before a tool acts, that Felspar decided on this case, that the released pack lets the outcome do this, and that the text is what Felspar checked.
- [Verify receipts](https://docs.felspar.ai/guides/verify-receipts): Confirm a Felspar decision with one REST call before your code acts on it.
- [Roles and permissions](https://docs.felspar.ai/guides/roles-and-permissions): Workspaces, owners, approvers, members and agents, what each can do, and how people join.

## Integrations

- [Overview](https://docs.felspar.ai/integrations): Every way to connect to Felspar - for people who author policy and for agents that use it - and which to choose.
- [ChatGPT](https://docs.felspar.ai/integrations/chatgpt): Use the Felspar plugin in ChatGPT to draft, check, propose, backtest and triage Decision Packs.
- [Claude](https://docs.felspar.ai/integrations/claude): Connect Felspar to claude.ai, Claude Cowork or Claude Code, for authoring as a person or for agents with a key.
- [Eve](https://docs.felspar.ai/integrations/eve): Connect an Eve agent to Felspar in two lines, install the Felspar skill, and gate the tools that act with the pack-driven action gate.
- [Claude Agent SDK](https://docs.felspar.ai/integrations/claude-agent-sdk): Connect an agent built with the Claude Agent SDK to Felspar, install the Felspar skill, and gate its acting tools with PreToolUse and PostToolUse hooks.
- [OpenAI Agents SDK](https://docs.felspar.ai/integrations/openai-agents-sdk): Connect an agent built with the OpenAI Agents SDK to Felspar, and gate its acting tools with tool guardrails driven by the released pack.
- [REST and n8n](https://docs.felspar.ai/integrations/rest-and-n8n): Call Felspar from code or an automation tool with three HTTP requests - brief, decision, verify - plus job lists and output checks.
- [Slack](https://docs.felspar.ai/integrations/slack): Felspar in Slack is not built yet. What is possible today, and what is coming.

## Reference

- [MCP tools](https://docs.felspar.ai/reference/mcp-tools): Felspar's remote MCP server - connection, the four agent tools and the seven authoring tools, with inputs, results and errors.
- [REST API](https://docs.felspar.ai/reference/rest-api): Felspar's REST routes - authentication, the agent routes, authoring and review routes, the decision result, and request formats.
- [Refusal codes](https://docs.felspar.ai/reference/refusal-codes): Every code Felspar returns when it refuses a request or can't read evidence, what it means and how to fix it.
- [Pack draft format](https://docs.felspar.ai/reference/pack-draft-format): The felspar.pack-draft JSON format that check_pack and propose_pack accept - purpose, subjects, questions, outcomes, escalations and test cases - with every field and limit.
- [Limits](https://docs.felspar.ai/reference/limits): Request sizes, rate limits, classification bounds, attempts, pack sizes and how long things are kept.

## Trust

- [Security, privacy and data](https://docs.felspar.ai/security): What Felspar stores and for how long, how workspaces are isolated, how access is checked and revoked, and where models are used.
- [FAQ](https://docs.felspar.ai/faq): Short answers to common questions about Felspar.
- [Glossary](https://docs.felspar.ai/glossary): Every Felspar term in one place, with a one-line definition.

## Agent skills

- [felspar-docs](https://docs.felspar.ai/.well-known/agent-skills/felspar-docs/SKILL.md): Read the Felspar Docs instead of answering from memory. Felspar turns your company's policy documents into tested, approved Decision Packs that every agent can share. Use when a task involves Felspar Docs: how it works, how to set it up or configure it, or its API.

## Agent resources

- [llms-full.txt](https://docs.felspar.ai/llms-full.txt): The full Markdown of every page in one file.
- [Page Markdown](https://docs.felspar.ai/index.md): Append `.md` to any page URL to fetch that page as raw Markdown.
- [JSON API](https://docs.felspar.ai/api/docs/pages.json): Page index of the JSON docs API; each entry links the page's JSON and Markdown forms. Described by the OpenAPI document at https://docs.felspar.ai/openapi.json.
- [Agent skills](https://docs.felspar.ai/.well-known/agent-skills/index.json): Agent Skills discovery index of the skills this site publishes.
- [API catalog](https://docs.felspar.ai/.well-known/api-catalog): RFC 9727 linkset of the APIs documented here.
- [AI catalog](https://docs.felspar.ai/.well-known/ai-catalog.json): ARD manifest of the agent-facing resources on this site (MCP server, skills, APIs).
- [agent-readability.json](https://docs.felspar.ai/agent-readability.json): Manifest of every agent-facing artifact on this site.
- [Sitemap](https://docs.felspar.ai/sitemap.xml): Every indexable page URL with its last-modified date.