Claude Pathmode connectorWhat Claude can do with your Pathmode product specs.
The Claude Pathmode connector exposes 22 tools. 10 read your intents and workspace, 9 can write to them, 3 stay ambiguous for lack of a clear source. Below: what Claude looks up, what it changes, what needs a connected workspace, and what's still fuzzy.
Verified Trustpilot reviews · AI, automation & growth agency
What changes for a product team shipping with a coding agent
Pathmode stores your product decisions as "intents": objective, expected outcomes, constraints, edge cases, verification. Once the extension is connected, its tools cover reading those specs, pulling the workspace's strategy and constitution, and logging what the work uncovered, so a coding agent starts from the spec rather than a vague request.
Start from the right spec. Before tackling a ticket, get_current_intent or search_intents find the relevant intent, and get_agent_prompt turns it into a structured execution prompt.
See what's blocking. analyze_intent_graph surfaces the critical path, cycles and bottlenecks across the dependency graph, while get_intent_relations zooms in on one intent.
Keep a paper trail. create_evidence logs a bug or a user quote, link_evidence ties it to an intent, and record_implementation_finding records what the spec got wrong.
What Claude doesn't do here: it doesn't close out work shipped through a pull request, since Pathmode leaves that to the merge. Most of these tools need a connected Pathmode workspace; without a key, only the project's intent.md file is in play. Nothing runs in the background either. To chain tasks on a schedule, you need an automation tool; the Integrations hub covers that side.
The vocabulary in one minute
Five words that come up on this page.
- Connector
- What you add once to open a tool or service to Claude; here, an extension that runs on your own machine.
- Tool
- A named action the connector makes possible. Claude decides on its own which ones to call, and the directory lists them by name.
- Intent
- Pathmode's word for a product spec: an objective, observable outcomes, constraints and edge cases, all in one record.
- Approval
- The confirmation Claude asks for before an action that changes something, shown in the chat at the right moment.
- MCP
- The open standard that links connectors to Claude; it's the channel the assistant uses to talk to Pathmode.
Connect Pathmode to Claude in three steps
- 01
Find Pathmode
In Claude's settings, open Customize, then Connectors, and spot Pathmode. It's a local extension. On a Team or Enterprise workspace, an Owner or Primary Owner has to enable it for the organization first.
- 02
Add it
Click Connect on its row. Without a key, Pathmode runs in local mode on the project's intent.md file. If adding it goes wrong, use Disconnect and try again.
- 03
Review what you authorize
If a service shows an authorization screen, read it before accepting: that screen, not Claude, sets how far access goes. On Pathmode's side, connecting a team workspace goes through a key created in the app.
The 22 tools, grouped by what they do
Pathmode gives Claude 22 tools: 10 that read your account, 9 that change something in it, and 3 no official source describes.
Three groups: what Claude reads, what it can write, and what the source describes without saying whether anything gets written. Names stay in English.
- 10 read
- 9 write
- 3 not documented
What Claude reads (10)
10 toolsTen tools that look up intents, the graph, the workspace and evidence without changing them.
get_current_intent
Returns the intent in play: the first approved one, or failing that the most recent. It's the shortcut when you've lost track of which spec drives the current work.
list_intents
Shows every intent in the workspace, with an optional status filter: draft, validated, approved, shipped or verified. You see where each spec stands without opening the app.
get_intent
From an ID, brings back a full intent with every field. Claude then works from the full spec, not just its title.
get_intent_relations
Shows the dependency graph around one specific intent: what it depends on and what depends on it. Handy for judging what slips if this one does.
search_intents
Searches by keyword across the goals, objectives and expected outcomes of every intent. Useful when you know the topic but not the ID.
analyze_intent_graph
Combs through the whole dependency graph to surface the critical path, cycles and bottlenecks. It's the big-picture view, where the dependencies tool zooms in.
get_agent_prompt
Builds a structured execution prompt for an intent, the instructions a coding agent follows to implement it. Pathmode also feeds the latest handoff notes into it.
get_workspace
Returns the Pathmode workspace details: strategy, active products and constitution. That's the high-level context behind every spec, with the constitution holding the rules any implementation must respect.
get_constitution
Gives the workspace's mandatory constraint rules, the ones any implementation has to respect. Pathmode also grades deliveries against them, rule by rule.
query_evidence
Searches evidence items by product, type, severity or free text, and returns their IDs so you can tie them to an intent later. Think of it as the lookup step before any linking.
What Claude can write (9)
9 toolsNine tools that create or change intents, evidence, notes or files. No source describes their confirmation, so the general rule applies.
sync_context
Approval: see the ruleWrites the workspace's canonical context into the repo's CLAUDE.md or AGENTS.md, replacing the Pathmode-marked section in place without duplicating it.
update_intent_status
Approval: see the ruleMoves an intent one step along: draft, validated, approved, shipped, then verified. Pathmode reserves it for work that never goes through a pull request, and for moving shipped to verified.
log_implementation_note
Approval: see the ruleLeaves a handoff note for the next session. The newest notes show up in the next execution prompt, which is why each should stand on its own.
record_implementation_finding
Approval: see the ruleRecords what the spec wrongly assumed and what building revealed. Targeting a specific check flips it to failing, and the finding rides into future prompts until a human reconciles it.
create_intent
Approval: see the ruleCreates an intent in the Pathmode workspace without writing a file. Inside a repo, Pathmode recommends intent_save instead.
update_intent
Approval: see the ruleUpdates an existing intent's content, passing only the fields that need to change. The rest of the spec stays as it was.
create_evidence
Approval: see the ruleCreates an evidence item: a bug found, a user quote, a behavioral observation or a feature request.
link_evidence
Approval: see the ruleLinks evidence to an intent, or unlinks it, to keep the thread between a user problem and the planned solution.
intent_save
Approval: see the ruleSaves a spec to the repo's intent.md: the root file, otherwise the single file in the intent folder, otherwise a new root file. It also works in local mode, with no key.
Direction unclear (3)
3 toolsPathmode describes these three tools without saying whether they write anywhere or just return a result. We won't decide that for them.
export_context
According to Pathmode, this tool generates CLAUDE.md, AGENTS.md, .cursorrules or intent.md files from an intent or a product. No source says whether it writes those files to disk or just returns their contents.
verify_implementation
Pathmode describes it as grading an implementation against the spec, using the pull request whose diff it fetches itself, criterion by criterion with citations. No source says whether that grade gets stored in the workspace.
intent_export
Per the publisher, it exports a spec as .cursorrules, as a CLAUDE.md or AGENTS.md section, or as an outcomes rubric. Nothing says whether the export lands in a file or stays in the reply.
What Claude asks before acting
By default, Claude stops and asks for your go-ahead before every action it takes for you on an account or service. The request shows up in the chat, at the moment it matters.
No source describes a confirmation specific to the nine writing tools, or to the three whose direction is unclear. Pathmode does set its own guardrails: an agent never signs off on its own delivery, the pull request merge decides the shipped status, and a proposed correction waits for a signed-in person to accept it. Pathmode also says its instructions to the agent are guidance, not enforcement. On Team and Enterprise, Owners can restrict the connector for the whole organization.
Which plans it's available on
None of the 819 sheets in the official directory shows plan availability. Per connector, that information isn't published anywhere.
The general rule is published: desktop extensions install in Claude Desktop, and remote connectors are open on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner enables them for the organization before members use them. The current state shows on the Pathmode sheet in the official directory.
Where this connector stops
A connector isn't an automation. Claude calls these tools while it answers, so no intent changes status on its own.
The tool list is an observed floor, not a ceiling: an admin can open actions that appear on no public sheet. This connector is documented by its publisher alone, whose guide also describes several tools missing from the sheet; the rest of the family sits on the Claude connectors hub. Anthropic reminds you on every sheet that it doesn't vouch for how exposed tools behave, so only add publishers you trust.
Need help connecting Pathmode to Claude?
A person reads every message.