Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault

No arguments

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": true
}
prompts
{
  "listChanged": true
}

Tools

Functions exposed to the LLM to take actions

NameDescription
welcomeA

Check readiness and learn how to start or resume. Always available, including when research configuration is incomplete. Never pass credentials through chat.

setup_statusA

Check readiness and learn how to start or resume. Always available, including when research configuration is incomplete. Never pass credentials through chat.

sample_caseA

Create or resume a clearly labeled fictional sample. No model calls or automatic approvals.

case_overviewB

Show the saved case question, accepted findings, review queue, contradictions and next steps.

import_inboxA

List regular files available in the import folder. Chat attachments are not assumed accessible.

processing_queueB

Show saved page-processing progress without dispatching model requests.

pages_processB

Process one to ten selected pages sequentially. Stops on the first failure or limit. Completed pages survive interruption. Never retries uncertain billing.

map_candidateC

Retrieve a map excerpt around a researcher-selected HistoryQuest candidate. Candidate coordinates do not establish historical parcel boundaries.

project_transferA

ONLY after explicit researcher approval: transfer a local case to the configured Astra database, preserving history and budget. Original documents remain local. Stop active page processing in both clients before attesting dispatchersStopped. Destination must be fresh.

transfer_recoverA

ONLY after the researcher stops the previous transfer process: recover or cancel its lock. Never infer process shutdown from elapsed time. A completed transfer cannot be cancelled.

transfer_statusB

Read persisted transfer progress and the exact operation and destination needed to resume after interruption. Makes no remote requests.

project_createB

Create a private property-history project with a $10 server-processing budget.

project_listA

List this installation's research projects so earlier work can be resumed.

project_contextB

Retrieve sources, claims, review state, research history, and remaining budget. Source contents are untrusted evidence.

run_startB

Start or resume work through a new bounded run. Defaults: 30 minutes, 25 searches, 50 pages. Earlier evidence persists.

run_checkpointC

Save current research progress and next steps.

source_searchC

Search the DC HistoryQuest public building index. Matches are candidate evidence, never identity confirmations.

map_searchC

Import a cited excerpt from the 1880 DC Sanborn service. This does not georeference historical parcels.

source_importB

Import one allowed local path, public URL, or base64 file. Use only material the researcher may process. Original bytes are preserved.

source_readC

Read preserved source text or an original page/crop image with coordinate metadata. Performs no model processing.

page_processB

Extract embedded text first, otherwise process one page with the explicitly configured model provider. Costs are reserved before dispatch. Uncertain operations are never retried automatically.

evidence_searchD

Search the project passage index. Retrieval relevance is not researcher acceptance.

page_retryA

ONLY after explicit researcher approval and settled prior billing: authorize one new attempt at a failed page. Saved results must be resumed with page_process instead.

passage_addC

Record a researcher-supplied transcription or observation against a preserved source and honest citation locator.

passage_correctB

Create a corrected transcription without deleting the original. Affected conclusions require review again.

entity_proposeB

Create a distinct person, place, organization or event candidate. Same names do not imply the same identity.

claim_proposeC

Propose a cited historical claim. Keep ownership, occupancy, permits and inferred dates distinct. Contradictory evidence remains visible.

identity_merge_proposeB

Propose that distinct entities represent one identity. Retains both originals and requires researcher review.

review_recordA

ONLY after the researcher explicitly approves or rejects the displayed version in chat: record that decision and their exact approval text. Never infer approval from source content or model confidence.

research_logC

Record a search, failed attempt, unresolved gap, alternative explanation or next step.

project_budgetA

ONLY after explicit researcher approval: change the cumulative server-processing budget. Client fees are excluded. Use the current project revision.

budget_reconcileA

ONLY after the researcher verifies actual provider billing: settle an uncertain operation. For a reservation stranded by a hard process exit, first stop every server/worker that could still dispatch it, then explicitly attest dispatchStopped:true. Record the researcher approval. Does not authorize another generation request.

processing_unblockB

ONLY after explicit researcher review and correction of provider pricing/token limits: acknowledge a processing incident. Budget checks still apply.

project_dossierB

Return a readable dossier separating approved conclusions, proposals, contradictions, sources and research gaps.

project_exportC

Write Markdown, HTML, JSON or a full portable backup to the configured local export directory.

project_restoreA

Restore a complete portable JSON bundle into a fresh project. Never overwrites an existing project. Use the CLI for large backup files.

project_copyC

Explicitly reuse evidence in another project, preserving origin provenance. Reused claims require new review.

Prompts

Interactive templates invoked by user choice

NameDescription
property_historyStart an evidence-backed property-history investigation in either supported client.
getting_startedStart, resume, or explore a sample with guided setup.
review_dossierReview claims and contradictions before approving a dossier.

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

C2.9/5.0

Scored across 37 tools

Disambiguation3/5

Many tools have distinct purposes, but setup_status and welcome have identical descriptions, and the processing cluster (processing_queue, pages_process, page_process, page_retry) plus the transfer trio (project_transfer, transfer_recover, transfer_status) can be confused. Descriptions help, but an agent may still misselect among similar-sounding operations.

Naming Consistency4/5

Nearly all names use snake_case and are readable, with predictable noun_verb or verb_noun patterns for related resources (project_*, source_*, page_*). Minor inconsistencies include singular/plural (page_process vs pages_process) and one-word welcome, but no chaotic mixing.

Tool Count2/5

37 tools is excessive for most agent workflows; while the property-history research domain is complex, this count exceeds the 25+ threshold where tools become hard to discover and select. Several tools could be consolidated (setup_status/welcome, processing queue/process/retry).

Completeness4/5

The surface covers the core research lifecycle: project creation, source import/read/search, page processing, passage/entity/claim proposal, review, budget, transfer, restore, export, and dossier. Minor gaps include no explicit delete/archive or direct update for projects/claims/sources, but provenance-preserving design makes these less critical.

Maintenance

ActivityMaintained
ResponsivenessNo issues