# Maplith

Start with get_map_overview. The complete reference below lists all tool consents; your key sees only its permitted subset.

## MCP tools
| Tool | What it answers | Listed to |
|---|---|---|
| `correct_source_system` | Correct upload system grouping without removing map items. | a key raised above read-only; applies only for an autonomous key that reads the whole map |
| `answer_question` | Answer a question Maplith asked about the data: a column, a row key, a dialect or an original. | a key raised above read-only; the grain, schema_dialect and lineage_direction kinds apply only for an autonomous key that reads the whole map, and a field_identity answer waits for a person unless the key is autonomous |
| `update_finding_status` | Change a finding’s status, immediately or after approval. | a key raised above read-only |
| `configure_metric` | Start, change or stop tracking a grounded metric. | a key raised above read-only; it applies only for an autonomous key |
| `register_automation` | Record an automation and the map objects it reads or writes. | a key raised above read-only; it applies only for an autonomous key |
| `take_signal` | Take a signal on a lease, or plan it within the workspace’s limits. | a key raised above read-only |
| `record_action` | Record work and its outcome; close or hand over a signal. | a key raised above read-only; a legacy key with an action policy, for the outcome of its own requests |
| `state_relationships` | State cross-table relationships for future scans to test. | an autonomous key that reads the whole map, never a published demo key or reviewer’s key |
| `draft_playbook` | Draft a playbook for approval, or record a person’s stated manual time. | a key that acts with approval or autonomously and reads the whole map, for the steps part; the whole part only a key of Maplith’s own workspace or the directory review sandbox’s reviewer; never a published demo key or a reviewer’s key outside the directory review sandbox |
| `suggest_feature` | Tell Maplith what its MCP server lacks or gets wrong, without customer data. | any key outside a demonstration workspace, and the directory review sandbox’s reviewer (recorded there, not filed), where Maplith has feature suggestions configured; never a new read-only connection |
| `join_review` | Question or add a join, or report counts; a person decides, unless an acting key's counts agree. | every key but the published demo key and a reviewer’s key outside the directory review sandbox; never a new read-only connection |
| `list_requests` | Read the decisions and status of your key’s action requests. | every key that can call suggest_feature, and every key the action gate is listed to; requests is filled only for a key that acts with approval or autonomously or holds an action policy, while waiting_on_maplith and recently_done are filled for every key that sees it; never a published demo key |
| `request_action` | Ask permission for an operation, or suggest a rule or live check for a person. | a key that acts with approval or autonomously, and a read-only key with an action policy on legacy connections only; never a published demo key or reviewer’s key |
| `redeem_action_grant` | Spend a single-use signed grant for an allowed action. | a key that acts with approval or autonomously, and a read-only key with an action policy on legacy connections only; never a published demo key or reviewer’s key; listed only where Maplith signs grants, and it redeems only once the key holds a policy |
| `report_observation` | Report observations for a live check an admin bound to this key. | a key an admin bound to a live check, including a read-only connection |
| `check_setup` | Check setup gaps and the next steps that repair them. | every key outside a demonstration workspace, and the directory review sandbox’s reviewer |
| `get_map_overview` | Orient yourself: map coverage, freshness, counts and next setup steps. | every key |
| `get_object` | Read an object’s applicable rules, fields, relations and caveats. | every key |
| `search` | Find map items and the decision rules that govern them. | every key |
| `fetch` | Fetch a map item by search reference or stable finding key. | every key |
| `get_process` | Read one process in full, or list them all when no object_type is given. | every key |
| `get_metric` | Read one metric in full, or list them all when no name is given. | every key |
| `list_findings` | List documented data problems with evidence and recommendations. | every key |
| `list_sources` | List the systems and datasets the current map came from. | every key |
| `list_signals` | Read the scoped work queue, lifecycle and counts. | every key |
| `list_automations` | List registered automations and the map objects they use. | every key |
| `get_task_context` | Start any work here: rules, context and playbooks for a task, a signal or a support case. | every key |
| `get_value_summary` | Summarise work and its stated value over a period. | every key |
| `route` | Find a verified route between objects, with join keys and caveats. | every key |
| `ask` | Answer a question from the map or Maplith’s documentation, with sources. | every key |
| `submit_sources` | Upload synthetic or customer-approved exports; append files with upload_id. | a key with uploads allowed (legacy consent or a writing policy), never in a demonstration workspace but the directory review sandbox |
| `build_map` | Start a scan from a submitted upload and follow its run. | a key with uploads allowed (legacy consent or a writing policy), never in a demonstration workspace but the directory review sandbox |
| `set_wake_subscription` | Create, change or test your key’s signed wake endpoint and filter. | an autonomous key, where wake-ups are on |
| `remove_wake_subscription` | Remove one of your key’s wake endpoints and revoke its secret. | an autonomous key, where wake-ups are on |
| `get_rule` | List the rules your key may see, or read one with its live checks’ reporting contract. | a key, where live rules are on for the workspace (in a demonstration workspace, a key bound to a live check) |

> Maplith builds a verified data-ontology map of a company's business systems
> — objects, fields (each with a personal-data classification set by the scan
> from the column name and the shape of its values, corrected by the customer),
> named and directed cross-system relations — each with a verb, a cardinality
> and, where the join could be tested against the data, an executable join spec
> and a verification rate — processes, grounded metrics (proposed by the map,
> or defined by an admin in the same grammar),
> and documented discrepancies (findings). The map is metadata (Maplith stores
> structure, not rows; findings quote evidence) and is served to AI agents over
> MCP with scoped keys and per-key write policies.

A scan also writes what changed as typed signals for an agent to poll (`list_signals`,
`get_task_context`): findings that appeared, process steps that slowed,
a tracked number that crossed the line the customer drew, a dependency that broke — and the
conditions an admin writes as rules over the map, judged per group as every scan finishes:
`rule_fired` when a group matches, `rule_broken` when the map can no longer answer the rule,
each answered once it no longer holds (a live rule's sender reports that as `rule_cleared`).
A rule's first evaluation is its baseline and fires nothing:
the groups that already hold are recorded and raise no signal, so a group fires when it
starts holding, or stops holding and later holds again. A signal is raised only when there is
work; news (a finding that went away, a rollback of the map) stays in the map's own records.
Every signal has a lifecycle: open, planned (picked up for later), in progress (an agent takes it with `take_signal` on a
lease, or it waits on a person) and done, with what came of it — `fixed`, `not_an_issue` or
`wont_fix`, reported with `record_action`. A fix a later scan still sees opens again with its
history; a decision to leave it stays quiet until it changes.

Maplith does not replace the agent that reads this. It hands that agent what it would otherwise
rebuild from zero on every task — where the data lives and how the systems join (measured against
the rows, not guessed), what needs doing (signals), what one task touches and how the business
handles it (`get_task_context`, with the playbooks the customer wrote or approved), and whether it
may act (`request_action`: allowed, pending a person, or denied; an allowed answer can carry a signed grant good once
for the exact target and parameters). The agent does the work in the customer's own systems with its own
credentials; Maplith never performs the action. Also on the same endpoint: `ask` (one question in plain
language, answered from the map the key may see, with citations and, for each join it cites, the
measured rate or that it was never measured), `route` (the verified join path between two objects), `list_sources` (the systems and files the map was built from, with format, record count and when each was last read), `list_automations` (the automations the business already runs, bound to the objects they read and write, the ones that never ran or last failed first), `get_value_summary` (counts of what
came of the signals, never an estimate), playbooks as MCP resources, prompts and a SKILL.md, and
Maplith's own documentation as MCP resources (maplith://docs/<page>) and prompts, and Maplith's own product map
(maplith://product-map): its concepts, how they connect, what each needs first, and the tool or page that serves each.
Confirmed glossary terms and joins never change on a rescan, whether a person or Maplith's build confirmed them, and every answer says who confirmed each. Where a scan disagrees, or would remove something, it raises a proposal for a person.
Before relying on the map, the agent tests it with its human: realistic scenarios per process and object, each hop's
evidence read, and every wrong or missing join flagged with `join_review`; the agent reports the counts that settle it
with the same tool, never a row. It changes nothing on the map unless a key that may act reports agreeing counts on a
review Maplith opened, which confirms that join in the glossary as tested by the customer's agent; otherwise a person decides.
Live rules — Maplith compiles what to check live, the customer's own automation observes and reports back,
and the observation becomes a signal within minutes — are in early access and not generally available.

MCP endpoint: https://app.maplith.com/api/mcp (streamable HTTP; Bearer agent key, or OAuth — discovery, dynamic client registration and consent, which issues the key for you).
Setting up with your own agent takes three steps (sign in, connect the agent, paste one setup prompt); the prompt builds the first map with your human, sets up the first actions and works the signals inside the action gate: https://maplith.com/docs/md/own-agent
Public read-only sandbox key: mlk_demo_acme-ltyWYChCrYzO8KOec-8BX-yO (fictional data, rate-limited).
Registry: com.maplith/maplith in the official MCP Registry.
Discovery: how to connect at https://maplith.com/auth.md, the MCP server card at https://app.maplith.com/api/mcp/server-card, the API catalog at https://maplith.com/.well-known/api-catalog, the Maplith skills at https://maplith.com/.well-known/agent-skills/index.json.
A page with a markdown twin (the docs, system guides, solutions, industries and legal pages, and the site pages below) answers with it at its own address when the Accept header prefers text/markdown.
When a call fails, the answer says what to do next: a JSON-RPC error carries `error.data.next_step`, and a tool that refuses
answers with a result marked `isError` whose `error` says why and whose `next_step` names the call to make, whether to wait, or who decides.

## What is a company brain?

A company brain is the knowledge a business runs on, in people's heads and across its systems, made available to the AI agents the company already uses.
Maplith builds one from the systems a company already uses: the map with measured joins, findings, playbooks, signals, task context, and the action gate where a person approves.
Maplith keeps metadata about a company's data, not copies of its records. It is not enterprise search, a chatbot over documents or a data warehouse.

- The article: https://maplith.com/docs/md/what-is-a-company-brain

## Why an agent should use Maplith

Maplith does not do your work; it gives you the part of every task you would otherwise rebuild from zero.

- Where the data lives and how the systems join: joins measured against the rows. In our test company a model asked to guess produced 4 complete joins; measurement finds 10 pairs that meet the rule, and the map publishes 8 (the 4 among them).
- Fewer tokens: about 6,154 tokens across three map calls per question, against about 26,851 tokens of raw exports every session, measured on an earlier synthetic test company; map answers are sized by the map, not by the number of rows.
- What needs doing (`list_signals`), one call for a task (`get_task_context`), how the business handles it (playbooks), and whether you may act (`request_action`).
- Maplith never returns rows, never performs the action it decides on and never holds the credentials you act with. Live rules are early access, not yet available.
- The full case: https://maplith.com/docs/md/why-agents-use-maplith

## Site pages

The site's own pages as markdown, generated from the pages.

- [Home](https://maplith.com/md/home): https://maplith.com
- [Product](https://maplith.com/md/product): https://maplith.com/product
- [Agents](https://maplith.com/md/agents): https://maplith.com/agents
- [Security](https://maplith.com/md/security): https://maplith.com/security
- [Pricing](https://maplith.com/md/pricing): https://maplith.com/pricing
- [About](https://maplith.com/md/about): https://maplith.com/about
- [Industries](https://maplith.com/md/industries): https://maplith.com/industries
- [Integrations](https://maplith.com/md/integrations): https://maplith.com/integrations
- [Docs](https://maplith.com/md/docs): https://maplith.com/docs
- [Solutions](https://maplith.com/md/solutions): https://maplith.com/solutions
- [The demo](https://maplith.com/md/demo): https://maplith.com/demo
- [The team](https://maplith.com/md/team): https://maplith.com/team
- [Support](https://maplith.com/md/support): https://maplith.com/support
- [Open source](https://maplith.com/md/open-source): https://maplith.com/open-source
- [Maplith Bridge updates](https://maplith.com/md/bridge-updates): https://maplith.com/bridge/updates
- [Cookies](https://maplith.com/md/cookies): https://maplith.com/cookies

## Docs

- [What is a company brain?](https://maplith.com/docs/md/what-is-a-company-brain): What the term means, why AI agents fail without one, what it is not, and how Maplith builds one from the systems a company already uses.
- [Getting started](https://maplith.com/docs/md/getting-started): From export files to a living map: upload, scan, explore.
- [MCP reference](https://maplith.com/docs/md/mcp): The tools, how to connect from Claude Code, claude.ai or ChatGPT, and the sandbox.
- [Agent keys & policies](https://maplith.com/docs/md/agent-keys): Scoped keys, autonomy policies, the hosted GitHub agent, and the audit trail.
- [Set Maplith up with your own agent](https://maplith.com/docs/md/own-agent): Three steps from no account to your own agent at work: sign in, connect it, paste one prompt. It builds your first map with you, sets up its first actions, works your signals and tells Maplith what is missing.
- [Signals & rules](https://maplith.com/docs/md/signals-and-rules): The pulse your agents poll, the rules you write over your own map, and why a new rule is quiet at first.
- [Wake your agent](https://maplith.com/docs/md/wake-your-agent): Maplith wakes your agent when something in your business moves: a signed call to your endpoint, how to verify it, the limits you set and how to start your own Claude agent.
- [Serve your data from your own network](https://maplith.com/docs/md/mcp-template): Run Maplith Bridge next to your exports: Maplith reads the files you choose, on your schedule, and never holds a password to your systems.
- [The GitHub connector](https://maplith.com/docs/md/github): Install the Maplith Agents app: the permissions it asks for and why, what it reads and keeps, the signal a delivery an agent acts on raises, how deliveries stay in your workspace, and how to disconnect.
- [Security & access model](https://maplith.com/docs/md/security): Metadata, not rows — and what can still appear in it, stated plainly; PII flags, RBAC in two dimensions, EU hosting with US model inference.
- [How pricing works](https://maplith.com/docs/md/pricing): Maplith is free to use today, within fair-use caps. Paid plans come later and will be announced in advance.
- [For agents](https://maplith.com/docs/md/for-agents): You are an AI agent? Start here: what Maplith gives you and how to use it well.
- [Why an agent should use Maplith](https://maplith.com/docs/md/why-agents-use-maplith): Written for AI agents: what you would otherwise rebuild from zero on every task, what Maplith hands you instead — measured — what it will never do, and when not to use it.

## System recipes

How to read each system into Maplith: the best path, whether the vendor runs an official MCP server, the datasets that make a good map, the read-only access to choose and the export steps. The same text is the MCP resource maplith://systems/<id>.

- [All system recipes](https://maplith.com/docs/md/systems)
- [Shopify](https://maplith.com/docs/md/systems/shopify)
- [WooCommerce](https://maplith.com/docs/md/systems/woocommerce)
- [e-conomic](https://maplith.com/docs/md/systems/e-conomic)
- [Dinero](https://maplith.com/docs/md/systems/dinero)
- [Billy](https://maplith.com/docs/md/systems/billy)
- [HubSpot](https://maplith.com/docs/md/systems/hubspot)
- [Pipedrive](https://maplith.com/docs/md/systems/pipedrive)
- [Intercom](https://maplith.com/docs/md/systems/intercom)
- [Zendesk](https://maplith.com/docs/md/systems/zendesk)
- [Stripe](https://maplith.com/docs/md/systems/stripe)
- [Meta Ads](https://maplith.com/docs/md/systems/meta-ads)
- [Google Sheets](https://maplith.com/docs/md/systems/google-sheets)
- [Recharge](https://maplith.com/docs/md/systems/recharge)
- [Slack](https://maplith.com/docs/md/systems/slack)
- [GitHub](https://maplith.com/docs/md/systems/github)
- [Email domain health (DMARC)](https://maplith.com/docs/md/systems/email-domain-health)
- [Google Workspace](https://maplith.com/docs/md/systems/google-workspace)
- [Metricool](https://maplith.com/docs/md/systems/metricool)
- [Triple Whale](https://maplith.com/docs/md/systems/triplewhale)

## Integrations

What an agent does with each system once Maplith has mapped it: example flows, how it connects, what Maplith never reads, and a FAQ. Each system's setup guide is under System recipes.

- [Shopify + AI agents](https://maplith.com/integrations/md/shopify): Orders, payments, products and customers from your shop. Connect Shopify to your own agent today; Maplith maps it from there.
- [WooCommerce + AI agents](https://maplith.com/integrations/md/woocommerce): Orders, refunds, products and customers from your WordPress shop. Connect WooCommerce to your own agent today, or share its exports; Maplith maps it from there.
- [e-conomic + AI agents](https://maplith.com/integrations/md/e-conomic): Invoices, customers and ledger entries, so sales meet the books. Share e-conomic’s exports today; Maplith maps them from there.
- [Intercom + AI agents](https://maplith.com/integrations/md/intercom): Conversations and tickets: who asks for help, about what, and how fast. Maplith’s Intercom integration is built and rolling out. Until it reaches your workspace, connect Intercom to your own agent today; Maplith maps it from there.
- [Stripe + AI agents](https://maplith.com/integrations/md/stripe): Payments, customers, invoices, subscriptions, balance transactions and payouts. Maplith’s Stripe integration is built and switched on workspace by workspace: ask Maplith at support@maplith.com to switch it on for yours. Until then, connect Stripe to your own agent today; Maplith maps it from there.
- [Recharge + AI agents](https://maplith.com/integrations/md/recharge): Subscriptions and recurring charges, tied back to your shop orders. Share Recharge’s exports today; Maplith maps them from there.

## Solutions

By department: the workflows an agent runs with Maplith in one business function, each shown on fictional companies' exports and linked to the worked example it comes from.

- [Maplith for marketing](https://maplith.com/solutions/md/marketing): How the AI agent a marketing team already uses stops paying for what cannot be sold and sees where the discount settings break the policy the business wrote down — each workflow a before and after, shown on fictional companies’ exports.
- [Maplith for customer service](https://maplith.com/solutions/md/customer-service): How the AI agent a support team already uses answers by the business’s own policy, asks before it refunds and — with live rules, in early access — tells a customer about a late parcel before they ask; each workflow a before and after, shown on fictional companies’ exports.
- [Maplith for product and catalogue teams](https://maplith.com/solutions/md/product-catalogue): How the AI agent a catalogue team already uses sees why returns climb on a product, and where the same product goes by two codes or two prices — shown on fictional companies’ exports.
- [Maplith for finance](https://maplith.com/solutions/md/finance): How the AI agent a finance team already uses finds the invoice that never matched its quote, the invoice still waiting for payment and the refund nobody booked — each workflow a before and after, shown on fictional companies’ exports.
- [Maplith for operations and delivery](https://maplith.com/solutions/md/operations): How the AI agent an operations team already uses sees the work that was handed out with no delivery after it — shown on fictional companies’ exports, next to a purchase order that arrived after the shelf was empty.
- [Maplith for sales and CRM](https://maplith.com/solutions/md/sales): How the AI agent a sales team already uses sees one customer split across several records, and a CRM that is days behind the customer — shown on fictional companies’ exports.
- [Maplith for people and resourcing](https://maplith.com/solutions/md/people): How the AI agent that plans the work sees the rates kept in a spreadsheet and the assignments made in chat — each workflow a before and after, shown on a fictional company’s exports.

## Industries

The worked examples: one fictional company per page, several departments working together.

- [Maplith for e-commerce](https://maplith.com/industries/md/e-commerce): How the AI agent a web shop already uses handles the everyday exceptions with Maplith — sold out while the ads run, a late parcel, returns climbing on one product, a support case that needs the policy — each as a before and after, with the pieces the agent uses.

## Legal

- [Terms & Conditions](https://maplith.com/legal/md/terms): The agreement for using Maplith: the free first map, your data and your agents, how the agreement changes, and what happens to a workspace no one uses.
- [Data Processing Agreement](https://maplith.com/legal/md/dpa): GDPR Article 28 terms for the personal data Maplith processes on your behalf: the purposes, the minimum security level, sub-processors and transfers, and how the processing description may change.
- [Processing and security description](https://maplith.com/legal/md/processing): How Maplith processes personal data today, for each feature by the kind of processing it does: which kinds of data, from which kinds of sources, for which purposes, for how long, with which sub-processors, and the security measures in force.
- [Sub-processors](https://maplith.com/legal/md/sub-processors): Which providers receive customer data, for what purpose, where they process it, and how we give notice before that changes.
- [Privacy policy](https://maplith.com/legal/md/privacy): What Maplith collects, why and on what legal basis, where it lives, what leaves the EU, how long it is kept, and how to exercise your rights.
- [Cookies](https://maplith.com/md/cookies): what Maplith stores in your browser, and why.

## Optional

- [Full docs in one file](https://maplith.com/llms-full.txt)
- [For agents](https://maplith.com/agents): the MCP surface explained for buyers
- [The demo](https://maplith.com/demo): the 46-second demo video, an AI agent working from the map of a fictional web shop, from a finding to a person's approval and the record
- [Security](https://maplith.com/security): the security model in depth
- [Integrations](https://maplith.com/integrations): every way Maplith reads a company's systems (GitHub, MCP servers, Maplith Bridge, the customer's own assistant, export files), what each reads, the access it asks for and why, the signals it raises, how to switch it off, and what is on the roadmap
- [The team](https://maplith.com/team): one founder and the AI agents that run Maplith with him, and what each one does
- [Support](https://maplith.com/support): how to get help — support@maplith.com, the docs, and the security contact
- [Open source](https://maplith.com/open-source): Maplith Bridge (https://github.com/Maplith/agents) and Maplith Wake (https://github.com/Maplith/agents/tree/main/wake-relay), Apache 2.0; the parts that run next to the customer's data and hold its keys, in the customer's own account
- [Maplith Bridge updates](https://maplith.com/bridge/updates): security notices and release news for Maplith Bridge by email, opt-in. Give the link to your person; never sign anyone up yourself.