Tools & skills

Once your client is connected, the Honeybase MCP server advertises its tools through the protocol's own tools/list. That list is the authoritative, always-current inventory — the best way to see exactly what your token can call is to have your client list its tools. This page is a map of the categories, not an exhaustive index.

Which tools appear depends on the scopes you approved and on your role: read tools need read, mutations need write, publishing needs release, and the organization-database tools need the orgdb.* scopes. Tools your grant or your deployment cannot use are simply not listed.

Tool categories

Area	What the tools do
Processes & workflows	List processes with their released and draft workflows, read a workflow's graph or a single node, read node-type documentation, save a workflow draft, duplicate a release into a fresh draft, delete an unreleased draft, and release a version
Runs	List a process's runs, read a run's result, and mint a short-lived download URL for a file a run produced. An assistant cannot start a run without your consent: request_run_authorization hands it a link you open in Honeybase to review what would run and confirm it — the run starts under your session — and get_run_authorization is how it learns whether you confirmed or declined. Choosing Confirm and always allow on that page gives a standing approval: later requests start at once and the result already says confirmed; revoke it any time from Barry's Standing approvals
Tasks	List and get tasks, and update one or many (assignee, due date, priority, team, notes, metadata, completion)
Teams & users	List the organization's users and teams, create and update teams
Schedules	List, create, update and delete job schedules
Availability & reassignment	Read availability and its history, set your own (or, with permission, others') availability, toggle a process as accepting, and plan or apply a reassignment
Agents & tools	List and get AI agents and their tools and runs, create and update agent drafts and tools, release an agent, pause or resume one
Forms	List and get the organization's reusable form library, validate a form schema, create and update forms
Organize	Folders and tags over the process / integration / agent / tool list views
Integrations & webhooks	List connected integrations and the live webhook ingest URLs
Organization database	Read the schema, run queries, write rows, and (separately gated) run DDL — behind the orgdb.* scopes
Dashboards	List the sources a chart can query (Honeybase's reporting views, the organization database, enabled SQL connections), describe their tables, look up column values, run a read-only query, check a chart config, save and edit charts, create and edit dashboards, and record what a table or column means — behind the dashboards.* scopes. They are the same tools Barry uses, under the same names; saves and edits happen at once, and deleting a chart or a dashboard is a request you confirm in the app

Process settings — timeouts, loop caps, failure notifications and JS packages — are also editable through a dedicated tool. For the vocabulary behind all of this (processes, workflows, runs, tasks, agents), see Concepts.

Everything an MCP token can do maps onto a permission you already have in the app; the MCP server never grants a new capability, it just exposes the existing ones to an AI client. Reads and write-echoes that include human-authored content are fenced as untrusted data before they reach the model.

The three skill-delivery tools

Three tools do not touch your organization's data at all. Instead they hand an agent the knowledge that lets it get things right the first time, since Honeybase's conventions are not in any model's training data. The first two return install instructions and download URLs for an authoring skill; the third returns its guide inline.

• get_graphql_api_skill — returns the download URLs and install steps for the Honeybase GraphQL API skill (a SKILL.md plus the full schema), so an agent can install it under .claude/skills/ and query the GraphQL API directly with an organization API key (hbk_...). Note the pairing: the skill is about calling the API, which uses an API key — distinct from the OAuth token the MCP connection itself runs on.

• get_forms_skill — returns the download URLs and install steps for the Honeybase Forms skill (a SKILL.md plus a SurveyJS property reference subsetted to exactly what Honeybase accepts), so an agent authors a form schema that both validates and renders correctly on the first attempt. The create_form and update_form tools point first-time authors here.

• get_dashboards_skill — returns, inline, how charts and dashboards are built: the loop (explore, query, check, save), {{name}} parameters and optional [[ ... ]] clauses, the reporting views, the chart config with an example per chart type, and every dashboard edit op. The dashboards tools point here before the first chart.

The two install tools return the hosted version, so an already-installed copy can be checked for updates. get_forms_skill and get_dashboards_skill are callable with any scope; get_graphql_api_skill needs read. They are the intended on-ramp when you want your agent building against Honeybase rather than only operating it.