Possibility Echo Agent bronze

Echo

Echoes back the input message for testing A2A connectivity

• Hello, agent!
• Echo this message back to me

MCP: apis

Public API catalog: list, search, and read full API definitions

MCP: capabilities

Public capabilities catalog: list, search, and read full capability definitions

MCP: examples

Public examples catalog: list, search, and read full example content

MCP: guides

Public developer guides catalog: list, search, and read full guide content

MCP: knowledge-base

Public knowledge-base catalog: list, search, read architectural and operational entries

MCP: mcp-servers

Public MCP servers catalog: list, search, and read full MCP server definitions

MCP: public-api

Unauthenticated tools: schema discovery, docs, validation

MCP: quickstarts

Public quickstarts catalog: list, search, and read full quickstart content

MCP: reference-bundles

Public reference bundles catalog: list, search, read curated agent-facing bundles

MCP: registries

Public registries catalog: list, search, read canonical pspace registry entities

MCP: rfcs

Public RFC catalog: list, search, and read full RFC content

Get the organizer-scoped A2A agent card

Get the organizer-scoped A2A agent card

• Invoke GET /.well-known/agent/organizer/{possibilityId}.json (mcp tool: get_organizer_a2a_agent_card)

Get the platform-admin A2A agent card

Get the platform-admin A2A agent card

• Invoke GET /.well-known/agent/admin.json (mcp tool: get_admin_a2a_agent_card)

List activities

List activities

• Invoke GET /admin/activities (mcp tool: admin_list_activities)

Get activity by ID

Get activity by ID

• Invoke GET /admin/activities/{id} (mcp tool: admin_get_activity)

Get activities by actor

Get activities by actor

• Invoke GET /admin/activities/actor/{xri} (mcp tool: admin_get_activities_by_actor)

List all corpora

Returns all analysis corpora. A corpus defines a collection of source materials (GitHub repos, branches, file paths) to analyze. Each corpus tracks import and conversion status, file counts, and is scoped to a Possibility.

• Invoke GET /corpora (mcp tool: list_corpora)

Create a new corpus

Creates an analysis corpus defining source material to analyze. Requires a name and GitHub source configuration (repo URL, branch, include/exclude paths). The corpus must be imported and transformed before analysis jobs can run against it.

• Invoke POST /corpora (mcp tool: create_corpus)

Get corpus by ID

Returns full details of a corpus including its source configuration, import/conversion status, file statistics, and associated possibility.

• Invoke GET /corpora/{id} (mcp tool: get_corpus)

Import corpus content

Triggers import of source files from the configured GitHub repositories into object storage. Runs asynchronously — poll get_import_status for progress. Must complete before corpus can be transformed or analyzed.

• Invoke POST /corpora/{id}/import (mcp tool: import_corpus)

Get import status

Returns the current import progress including files discovered, files imported, any errors, and overall completion status (pending, running, completed, failed).

• Invoke GET /corpora/{id}/import-status (mcp tool: get_import_status)

List corpus files

Returns the list of files imported into a corpus, including file paths, sizes, and types. Use after import completes to inspect what source material is available for analysis.

• Invoke GET /corpora/{id}/files (mcp tool: list_corpus_files)

Get corpus file source content

Retrieves the raw source content of a specific file within the corpus. The file path is specified as a URL path parameter. Returns the original content as imported from the source repository.

• Invoke GET /corpora/{id}/files/source/{path} (mcp tool: get_corpus_file_source)

Check corpus staleness

Compares the playbook's source snapshot (commit SHAs at synthesis time) against current repository commits. Returns whether the playbook is stale and which sources have changed since last synthesis.

• Invoke GET /corpora/{id}/staleness (mcp tool: check_corpus_staleness)

Get playbook history for corpus

Returns version history metadata for the playbook generated from this corpus, including version numbers, creation/modification dates, changelog entries with triggers and summaries, and provenance (source repos, commits, AI model used).

• Invoke GET /corpora/{id}/playbook-history (mcp tool: get_playbook_history)

Transform corpus

Converts raw imported files (markdown, D2 diagrams, YAML, instruction files) into .kno entities. Runs converters for each file type and writes output to Hive storage with change detection. Must complete before analysis jobs can use the corpus.

• Invoke POST /corpora/{id}/transform (mcp tool: transform_corpus)

List all evaluations

List all evaluations

• Invoke GET /evals (mcp tool: admin_list_evals)

Create a new evaluation

Create a new evaluation

• Invoke POST /evals (mcp tool: admin_create_eval)

Get evaluation by ID

Get evaluation by ID

• Invoke GET /evals/{id} (mcp tool: admin_get_eval)

Run an evaluation

Run an evaluation

• Invoke POST /evals/{id}/run (mcp tool: admin_run_eval)

List evaluation runs

List evaluation runs

• Invoke GET /evals/{id}/runs (mcp tool: admin_list_eval_runs)

Get evaluation run by ID

Get evaluation run by ID

• Invoke GET /eval-runs/{id} (mcp tool: admin_get_eval_run)

List all jobs

Returns all analysis jobs. An analysis job is a multi-phase pipeline that runs dimensional analysis (across 10 analytical lenses) on a corpus, then synthesizes results into a playbook. Each job tracks status, progress, and phase breakdown.

• Invoke GET /jobs (mcp tool: list_analysis_jobs)

Create a new job

Creates an analysis job for a corpus. The corpus must have completed import and transformation. Specify which dimensions (1–10) to analyze and how to partition them into phases.

• Invoke POST /jobs (mcp tool: create_analysis_job)

Get job by ID

Returns full details of an analysis job including its status, phase configuration, weighted progress (75% dimensional analysis + 25% post-synthesis), associated corpus, and synthesis state.

• Invoke GET /jobs/{id} (mcp tool: get_analysis_job)

List job runs

Returns all dimension runs across all phases of a job. Each run represents one analytical dimension's execution, with status (pending, running, completed, failed), timing, and checkpoint information.

• Invoke GET /jobs/{id}/runs (mcp tool: list_job_runs)

Check job readiness

Pre-flight validation that checks object storage connectivity, corpus files in storage, AI service credentials, and job/corpus data integrity. Must pass before a job can be started. Returns detailed check results.

• Invoke GET /jobs/{id}/readiness (mcp tool: check_job_readiness)

Start a job

Initializes a job by creating dimension run records for each phase and setting the job to queued status. Does not execute analysis — use execute_analysis_job or execute phases individually after starting.

• Invoke POST /jobs/{id}/start (mcp tool: start_analysis_job)

Execute a job

Runs all phases of the job asynchronously with configurable batch size and max files. Processes each dimension through the analytical pipeline with checkpointing for resume capability. Returns immediately with execution status.

• Invoke POST /jobs/{id}/execute (mcp tool: execute_analysis_job)

Resume a paused job

Resumes a job that was paused or partially completed. Picks up from the last checkpoint, preserving completed dimension runs and continuing from where execution left off.

• Invoke POST /jobs/{id}/resume (mcp tool: resume_analysis_job)

Synthesize job results

Runs the 6-step synthesis pipeline on completed dimensional analysis: load results, run 10 contextual lens passes, cross-lens meta-synthesis, extract principles/procedures/patterns, assemble into a .kno playbook, and evaluate quality. Supports step-level resume.

• Invoke POST /jobs/{id}/synthesize (mcp tool: synthesize_job)

Evaluate a job

Re-runs the quality evaluation step of synthesis on a completed playbook. Resets the evaluate and promote steps, then re-executes them to produce updated quality scores and metrics.

• Invoke POST /jobs/{id}/evaluate (mcp tool: evaluate_job)

Generate components from job

Derives glossary and workflow .kno entities from a completed playbook. Extracts terms and procedures from the playbook content, generates individual .kno files for each, and writes them to the Hive output path. Requires job to be in completed status.

• Invoke POST /jobs/{id}/generate-components (mcp tool: generate_components)

List all phases

Returns all phases for an analysis job. Each phase groups a subset of the 10 analytical dimensions, has a phase number, and tracks its own execution status and dimension runs.

• Invoke GET /phases (mcp tool: list_analysis_phases)

Create a new phase

Creates a phase within an analysis job. Specify which dimensions (1–10) to assign, file filters, and phase ordering. Phases partition the analytical work so dimensions can be executed in batches.

• Invoke POST /phases (mcp tool: create_analysis_phase)

Get phase by ID

Returns full details of a phase including its assigned dimensions, file filters, execution status, and associated dimension run results.

• Invoke GET /phases/{id} (mcp tool: get_analysis_phase)

Start a phase

Initializes dimension run records for each dimension assigned to this phase and sets the phase to running status. Must be called before execute.

• Invoke POST /phases/{id}/start (mcp tool: start_analysis_phase)

Execute a phase

Runs all assigned dimensions asynchronously through the analytical pipeline with configurable batch size and max files per dimension. Each dimension processes corpus files through its analytical lens with checkpointing.

• Invoke POST /phases/{id}/execute (mcp tool: execute_analysis_phase)

Retry failed phase runs

Re-executes only the dimension runs that failed within this phase, preserving completed dimension results. Useful for recovering from transient errors without redoing successful work.

• Invoke POST /phases/{id}/retry-failed (mcp tool: retry_failed_phase)

List phase runs

Returns all dimension runs for a specific phase. Each run represents one dimension's execution with status (pending, running, completed, failed), timing data, and result references.

• Invoke GET /phases/{id}/runs (mcp tool: list_phase_runs)

Get a specific dimension run

Returns full details of a single dimension run including its analytical dimension number, execution status, timing, checkpoint data, and result references in object storage.

• Invoke GET /runs/{id} (mcp tool: get_phase_run)

Describe a capability's config surface — secret-store keys and non-secret config…

Derive the variables an installed capability declares in its own manifest: secret-store keys (presence-only, `generated: true` marks platform-minted) and non-secret config variables (`required`, `default`, `format: string|url|enum`, `enum`), each tagged `kind: secret | config`, with the atom's `path_template` and the capability's `required_permissions`. Non-conforming manifests degrade with a `degrade_reason`, never a 5xx.

• Invoke GET /admin/possibilities/{possibilityId}/capabilities/{capabilitySlug}/config (mcp tool: describe_capability_config)

Get config status — presence only for secrets, value included for non-secret con…

For every declared variable, whether a value is stored at the atom's resolved path. Secret-store keys return `set` only — no response shape carries a secret value. Non-secret config variables also return their `value` when set, plus `required` and `default`. `missing_count` counts unset user-supplied secrets and unset required config variables without a default. An unreachable store degrades the atom to `status: unknown`.

• Invoke GET /admin/possibilities/{possibilityId}/capabilities/{capabilitySlug}/config/status (mcp tool: get_capability_config_status)

Set a config variable — a user-supplied secret or a non-secret config value

Write `{ value }` for one variable the capability's manifest declares into the atom's declared path (read-merge-write; siblings survive). Secret-store keys without `generated: true` and every config variable are writable; a config value must satisfy its declared `format` (`url`: absolute http(s) URL; `enum`: one of `enum[]`). Undeclared key → 404; platform-minted key or format violation → 422; store unavailable → 503. The value is never logged and never echoed.

• Invoke PUT /admin/possibilities/{possibilityId}/capabilities/{capabilitySlug}/config/{key} (mcp tool: set_capability_config_variable)

Get external capability token info

Get external capability token info

• Invoke GET /admin/possibilities/{possibilityId}/capabilities/{capabilitySlug}/token (mcp tool: get_external_capability_token)

List installed capabilities

List installed capabilities

• Invoke GET /admin/possibilities/{possibilityId}/capabilities (mcp tool: list_installed_capabilities)

Get external artifact A2UI surface

Get external artifact A2UI surface

• Invoke GET /admin/possibilities/{possibilityId}/capabilities/{capabilitySlug}/surface/artifacts (mcp tool: get_external_capability_artifacts)

Initiate domain verification

Initiate domain verification

• Invoke POST / (mcp tool: initiate_publisher_verification)

List publisher domain verifications

List publisher domain verifications

• Invoke GET / (mcp tool: list_publisher_verifications)

Get domain verification status

Get domain verification status

• Invoke GET /:id (mcp tool: get_publisher_verification)

Trigger DNS verification check

Trigger DNS verification check

• Invoke POST /:id/check (mcp tool: check_publisher_verification)

Get DNS instructions for verification

Get DNS instructions for verification

• Invoke GET /:id/dns (mcp tool: get_publisher_verification_dns)

List a collection

Enumerate every public entity in `{collection}`. Pagination uses opaque cursors — pass `next_cursor` from one page as `cursor` on the next.

• Invoke GET /v1/{collection} (mcp tool: collections_list)