Project Telos
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
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
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| telos.statusA | Use when a host needs current Telos workbench readiness and next actions. Read-only, zero-auth, no external side effects. Returns a JSON action envelope. |
| telos.doctorA | Use before demos, listings, or agent runs to check local Telos operator-spine health. Read-only, zero-auth, no external side effects. Returns JSON check results. |
| telos.roomA | Use when an agent needs the current five-flagship room summary before routing work. Read-only and zero-auth. Requires the sibling gather, crucible, index, and forum source checkouts next to this repo plus a local python interpreter; without them it returns an UNVERIFIABLE envelope naming the missing dependency rather than a live summary. Returns a JSON action envelope. |
| telos.workflowA | Use when validating the local five-flagship golden workflow from source checkouts. Read-only and zero-auth, no external side effects beyond local subprocess reads. Requires the sibling gather, crucible, index, and forum source checkouts next to this repo plus a local python interpreter; without them it returns an UNVERIFIABLE envelope naming the missing dependency rather than running the workflow. Returns JSON receipts and verdict counts. |
| telos.catalogA | Use when a host needs the provider-neutral catalog of Project Telos MCP tools and next actions. Read-only, zero-auth, no external side effects. Returns a JSON catalog. |
| telos.server.manifestA | Use when configuring MCP clients for gather, index, forum, crucible, and telos source checkouts. Read-only, zero-auth, no external side effects. Returns a JSON server manifest. |
| telos.mcp.freshnessA | Use when a host must compare loaded MCP servers against expected versions, tools, and probes. Read-only, zero-auth, no external side effects. Returns JSON MATCH, DRIFT, or UNVERIFIABLE freshness receipts. |
| telos.ci.doctorA | Use when a host needs GitHub Actions runtime, action-major, and latest five-flagship CI compatibility receipts. Read-only, zero-auth, no external side effects. Returns a JSON CI doctor register. |
| telos.ci.triageA | Use when a host needs to separate fatal GitHub Actions gate failures from Node runtime migration warnings before routing remediation. Read-only, no external side effects, and no external writes. Supports zero-auth offline fixture packets and live gh run intake through the CLI using local gh auth when required. Returns JSON MATCH, DRIFT, or UNVERIFIABLE CI triage receipts. |
| telos.presentation.doctorB | Use when a host needs five-flagship README, changelog, and brand-asset presentation parity receipts. Read-only, zero-auth, no external side effects. Returns JSON MATCH, DRIFT, or UNVERIFIABLE presentation receipts. |
| telos.accessibility.doctorA | Use when a host needs static HTML accessibility, reduced-motion, keyboard, and canvas-fallback receipts for Telos Studio surfaces. Read-only, zero-auth, no external side effects. Returns JSON MATCH, DRIFT, or UNVERIFIABLE accessibility receipts. |
| telos.performance.doctorA | Use when a host needs static Studio performance, efficiency, asset-budget, and embedding receipts. Read-only, zero-auth, no external side effects. Returns JSON MATCH, DRIFT, or UNVERIFIABLE performance receipts. |
| telos.compatibility.doctorA | Use when a host needs CLI, MCP, protocol, manifest, and integration compatibility receipts. Read-only, zero-auth, no external side effects. Returns JSON MATCH, DRIFT, or UNVERIFIABLE compatibility receipts. |
| telos.operator.doctorA | Use when a host needs README, status, catalog, manifest, CI, and current-state discoverability receipts for the Telos operator surface. Read-only, zero-auth, no external side effects. Returns JSON MATCH, DRIFT, or UNVERIFIABLE operator receipts. |
| telos.admission.telemetryA | Use when designing trace fields that keep action admission separate from verification verdicts. Read-only, zero-auth, no external side effects. Returns a JSON telemetry convention. |
| telos.context.envelopeB | Use when large-workspace context needs readable source refs, budgets, and receipt chains. Read-only, zero-auth, no external side effects. Returns a JSON context-envelope convention. |
| telos.context.packA | Use when preparing a bounded handoff packet for large-codebase agent work. Read-only, zero-auth, no external side effects. Returns a validated JSON context pack with hashes and verdicts. |
| telos.action.receiptB | Use when modeling proposed action, admission, execution, review, and compensation records. Read-only, zero-auth, no external side effects. Returns a JSON action-receipt interface. |
| telos.loop.ledgerA | Use when an agent loop needs durable state across fresh contexts and bounded scheduled runs. Read-only, zero-auth, no external side effects. Returns a JSON loop-ledger convention. |
| telos.objective.monitorA | Use when checking whether proxy metrics are drifting away from real agent or build objectives. Read-only, zero-auth, no external side effects. Returns JSON objective-monitor signals. |
| telos.model.foundryA | Use when planning model-foundry work across hosted, local, open-weight, post-training, and verifier gates. Read-only, zero-auth, no external side effects. Returns a JSON model-foundry contract. |
| telos.learning.forgeB | Use when video, channel, paper, and benchmark leads need to become receipt-backed learning labs before synthesis. Read-only, zero-auth, no external side effects. Returns a JSON Learning Forge packet. |
| telos.learning.labsA | Use when Learning Forge source leads need executable lab contracts with measurements, failure cases, and five-flagship ownership. Read-only, zero-auth, no external side effects. Returns a JSON Learning Forge labs packet. |
| telos.research.seedB | Use when terse research notes need source-backed seed packets before synthesis. Read-only, zero-auth, no external side effects. Returns JSON research seeds with provenance status. |
| telos.research.thermodynamicA | Use when exploring thermodynamic or stochastic AI-chip research through public transcript evidence. Read-only, zero-auth, no external side effects. Returns a JSON research receipt with verification labels. |
| telos.rendering.researchA | Use when collecting rendering leads for clustered-forward, Gaussian splatting, creative coding, and graphics demos. Read-only, zero-auth, no external side effects. Returns JSON research seeds. |
| telos.rendering.capabilitiesA | Use when a host must choose WebGPU, WebGL, canvas, or static rendering fallbacks for Studio surfaces. Read-only, zero-auth, no external side effects. Returns a JSON renderer capability contract. |
| telos.measurement.layersB | Use when visual, splat, lighting, dither, audio, uncertainty, or frame-budget evidence needs meters, or when a caller's own image (raw 8-bit RGBA, inline base64 or a file under an allowed local root) needs Telos measurement layers L0 to L3 with SHA-256 receipts. Called with no arguments it returns the demo meters. Read-only, zero-auth, no external side effects. Returns a JSON measurement packet (v1 demo or v2 caller image). |
| telos.creative.engineA | Use when presenting the Telos Creative Engine across generative art, sound, typography, media, CGI, math, and physics lanes. Read-only, zero-auth, no external side effects. Returns a JSON creative-engine manifest. |
| telos.creative.kernelsA | Use when deterministic creative primitives are needed for dithering, pixel sorting, plotter paths, or clustered-light bins. Read-only, zero-auth, no external side effects. Returns JSON creative kernels. |
| telos.revival.registryA | Use when deciding which older, siloed, or frozen local tools should be promoted into flagship lanes. Read-only, zero-auth, no external side effects. Returns a JSON revival registry. |
| telos.second_level.queueA | Use when assessing public-safe second-level flagship candidates before registry promotion. Read-only, zero-auth, no external side effects. Returns a JSON second-level flagship queue. |
| telos.workstation.substrateB | Use when a host needs public-safe aggregate intake for local workstation repositories and private/local lane families. Read-only, zero-auth, no external side effects. Returns a JSON workstation substrate register. |
| telos.display.calibrationA | Use when display, color, ICC/LUT, artifact refs, or measurement gates need a non-mutating calibration contract. Read-only, zero-auth, no external side effects. Returns a JSON display-calibration contract. |
| telos.native.controlA | Use when a host needs the Telos native background-control capability catalog for browser (Chrome DevTools Protocol) and native-app (Windows UI Automation) actuation, plus the device and learn verbs. Read-only, zero-auth, no external side effects. Returns a JSON capability and verb catalog. |
| telos.browser.evidenceA | Use when a host needs the synthetic browser evidence fixture for contract inspection. Read-only, zero-auth, no external side effects. Returns a JSON browser-evidence packet marked unredacted; URL and title context is retained. |
| telos.showcase.scoutA | Use when a host needs fixture-backed OSS Proof Showcase candidate rankings before public patch work. Read-only, zero-auth, no external side effects. Returns JSON scout results. |
| telos.proofA | Use when a host needs the fixture-backed agent-action proof packet joining source refs, context refs, route, admission, side effects, output digests, verifier checks, and the Emet witness stage. Read-only, zero-auth, no external side effects beyond local subprocess reads. Returns a JSON agent-action proof packet. |
| telos.proof.researchA | Use when a host needs the fixture-backed research-claim proof packet joining source provenance, a bounded claim, a required negative control fixture, attempt records, recomputed source digests, and a promotion rung the verifier derives rather than trusts. Read-only, zero-auth, no external side effects beyond local subprocess reads. Returns a JSON research-claim proof packet. |
| telos.proof.visualA | Use when a host needs the fixture-backed visual-truth proof packet whose color and luminance measurements are recomputed from the artifact's own embedded sRGB samples, with a read-only boundary and no physical-calibration overclaim. Read-only, zero-auth, no external side effects beyond local subprocess reads. Returns a JSON visual-truth proof packet. |
| telos.proof.buildA | Use when a host needs the fixture-backed build scientific-runtime proof packet whose conserved-quantity invariant and conservation drift are recomputed from the run's own embedded samples with stdlib math, checked within bounded tolerances, and controlled by a required negative fixture that must break the invariant. Read-only, zero-auth, no external side effects beyond local subprocess reads. Returns a JSON build proof packet. |
| telos.reach.doctorA | Use before any reach call to see which channels (web crawler, Reddit, X, official APIs) work for this user and what each one costs and leaves out. Read-only, zero-auth, no external side effects: reads environment variable names only, with no network, file writes or processes. Returns a JSON doctor report. |
| telos.reach.menuA | Use when an agent needs the reach verbs its grant tier allows, with credentials checked and a memory of what was already read. Read-only, zero-auth, no external side effects; reads the memory file named by |
| telos.crawl.fetchA | Use when an agent needs the text, feed items, sitemap URLs or JSON of one public URL. Read-only and zero-auth; checks robots.txt first and stops at a bot check. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON with the parsed page and a receipt of every fetch. |
| telos.crawl.siteA | Use when an agent needs several pages of one site; sitemaps and feeds are read before links, inside a depth (0 to 3) and page budget (1 to 50). Read-only and zero-auth; obeys robots.txt and Crawl-delay. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON page summaries and a receipt. |
| telos.reddit.listingA | Use when an agent needs posts from a subreddit through Reddit's official Data API. Read-only; auth is the user's own Reddit app credentials from environment variables named by telos.reach.doctor, paced from Reddit's rate-limit headers. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON posts and a receipt. |
| telos.reddit.commentsA | Use when an agent needs one Reddit post and its comment tree through the official Data API. Read-only; auth is the user's own Reddit app credentials from environment variables named by telos.reach.doctor. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON comments with depth and a receipt. |
| telos.x.oembedA | Use when an agent needs the text, author and date of one public X post from its URL, through X's public oEmbed endpoint. Read-only and zero-auth; no search or timelines. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON and a receipt. |
| telos.x.apiA | Use when the user has an X developer key and wants a post lookup, recent search or a user's recent posts through the official X API v2. Read-only; auth is X_BEARER_TOKEN, and X bills each read, so the receipt carries a cost estimate. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON and a receipt. |
| telos.reach.apiA | Use when an agent needs one operation from an official API: hackernews, github, v2ex, wikipedia, arxiv (zero-auth), or youtube and brave (auth is the user's own key in YOUTUBE_API_KEY or BRAVE_SEARCH_API_KEY). Read-only; keys travel in headers and never in receipts. Reaches the network: one honest Telos User-Agent, per-host pacing with backoff on 429 and 503, no login, no cookies. Returns JSON and a receipt. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 50 tools
Each tool has a specific 'Use when' target, but with 50 tools across repeated families such as doctor, proof, research, context, and learning, boundaries blur. Choosing among telos.doctor, telos.operator.doctor, domain-specific doctors, and various proof/convention packets requires careful description reading, making misselection likely.
Names uniformly follow a lowercase dotted telos.<domain>.<noun> hierarchy with no camelCase/snake_case mixing. The convention is predictable throughout, even though it is noun-oriented rather than verb_noun.
50 tools is far beyond the typical 3-15 range and creates a heavy selection surface even for a broad Telos workbench. Many tools are narrow diagnostic, proof, or convention variants, so the count feels inflated rather than well-scoped.
The read-only Telos workbench surface covers many domains—status, diagnostics, proofs, reach/crawl, research, rendering, context, CI, accessibility—so it is broad and hard to call incomplete. However, it contains no mutating, promotion, or execution operations, and several workflows end at queues/registries/receipts, leaving minor dead ends.