TVIJO
RUCart 0

COMP · v2 · 2026-10-15

More than a star count.

Choose a ready assistant, a development framework or a scoped TVIJO implementation. The right answer starts with the outcome you need.

Need a personal assistant?

Inspect OpenClaw and Hermes for documented channels, memory and operator entry. Your permissions and task quality still need checking.

Building an application?

Inspect Agents SDK, LangGraph, CrewAI, n8n or Codex for the role they fit. A framework still needs domain data, acceptance and an owner.

Need an accepted business workflow?

Order a scoped TVIJO pilot: connect the data, worker roles, budget, review and operator handoff. Existing tools can be part of the design.

The general decision table

TVIJO-authored review,15 October 2026. Documented capabilities are separate from installation evidence. Proposed Hotido/FinEdg uses are not shipped integrations. Unknown does not mean absent.

Compare mechanisms, delivery and proposed fit. No aggregate winner.
Solution & documented optionDelivery / evidenceTVIJO: current scope and useful gapProposed domain use
OpenClaw

Channel gateway, agent routing and memory

Buying guide →

Documented personal-agent runtime

Primary documentation; not installed or benchmarked here.

TVIJO has operator/Telegram components; a comparable packaged multi-channel onboarding and channel-health experience is not established in this review.

Hotido → Offer a source-linked tender alert in a messaging channel.

FinEdg → Deliver a read-only research digest with explicit data age.

Hermes Agent

Persistent memory and reusable skills

Buying guide →

Documented assistant runtime

Primary documentation; not installed or benchmarked here.

TVIJO already uses skills and persistent project records. A buyer-facing memory correction, expiry and supersession flow still needs explicit evidence; skills are not absent.

Hotido → Reuse a supplier-checking procedure; correct an outdated identity note.

FinEdg → Reuse a research checklist with instrument units and source dates.

OpenAI Agents SDK

Agent handoffs, traces and evaluation datasets

Buying guide →

SDK; hosted evaluation is a separate dependency

Primary documentation; not installed or benchmarked here.

TVIJO has criteria/evidence/review components. A public end-to-end dataset-to-grade example is a packaging target, not proof those mechanisms are missing.

Hotido → Evaluate a tender relevance filter on labelled public examples.

FinEdg → Check research answers against sources and numerical references.

OpenAI Codex

Repository tools, resumable sessions, review and subagents

Buying guide →

Coding agent; account quota and availability apply

Primary documentation; not installed or benchmarked here.

Codex is already part of the operator workflow. It can be a worker inside the swarm rather than a mutually exclusive replacement for TVIJO. No like-for-like quality benchmark exists here.

Hotido → Develop a source connector with one owner per file set and parser fixtures.

FinEdg → Implement and review paper-research tests with no order permissions.

LangGraph

Explicit graph state, durable execution and human checkpoints

Buying guide →

Orchestration runtime; LangSmith is separate

Primary documentation; not installed or benchmarked here.

TVIJO has board/checkpoint code. A general visual graph editor and end-to-end recovery across all workers are not established. Checkpoint presence alone is not a recovery guarantee.

Hotido → Resume a source-normalisation job after interruption without duplicate notices.

FinEdg → Resume a recorded-data research pipeline; keep the order path disconnected.

CrewAI

Role-based crews separated from control flows

Buying guide →

Crew/flow framework; tracing and checkpoints documented

Primary documentation; not installed or benchmarked here.

TVIJO has role/task contracts and an executor ladder. The useful gap is a clear public role-output example and deterministic stop/review flow, not inventing roles again.

Hotido → Collector → verifier → digest writer, each with a declared output contract.

FinEdg → Researcher → calculation check → critic, with a fixed stopping rule.

n8n

Visual workflows, AI steps and Telegram node

Buying guide →

Workflow platform; hosting and license terms must be checked

Primary documentation; not installed or benchmarked here.

TVIJO has a custom operator interface and scoped Telegram service. A comparable general-purpose drag-and-drop workflow editor is not established here.

Hotido → Show a buyer a visible source → filter → review → alert prototype.

FinEdg → Connect a read-only report to a human review queue; no exchange keys.

Where TVIJO can be the better fit

For a buyer who needs an integrated, accepted deliverable and an operator handoff, the engagement can matter more than a packaged feature count. These are scoped strengths, not exclusive features or a measured victory.

Control the implementation and deployment

Our own project source lets us adapt task contracts, data handling and deployment to the engagement. Delivery and licensing rights are agreed in the contract.

Not unique ownership: several alternatives also expose source or support custom code. Source access does not prove speed, security or product quality.

Use a bounded team, not an agent-count slogan

Project components include task ownership, criteria, checkpoints and authenticated messages. Their scope is explicit. Codex/Claude subscriptions and DSHREM workers can have different roles.

Message-auth tests are scoped, not a universal worker rollout or OS sandbox. The1/3/6-worker experiment is a protocol, not a measured scaling win.

Inspect cost per accepted result

Metered DSHREM tasks have caps, holds and provider-reported charges. A cost worksheet includes failed work and review; subscription quota stays separate.

A budget cap is not a price quote or proof of cheaper delivery. Unknown spend remains unknown and blocks a final total.

Inspect COMP-v1: yellow stars and technical-coverage checks

Historical rubric, not a quality ranking.1=scoped/custom component,2=documented support; five scoped components produce5/10. Unknown totals remain unknown. No scores were increased to improve our position.

Choose products

JSONHow the scores work →

2 per check: documented capability ·1: scoped or custom building block ·0: explicit absence ·?: not established. Five checks make 10. Unknowns leave the total blank.

Editorial coverage by criterion; no overall product ranking
CriterionTVIJOScoped engineering services + project componentsOpenClawPersonal-agent runtimeHermes AgentPersonal-agent runtimeOpenAI Agents SDKSDK + optional evaluation servicesOpenAI CodexCoding agentLangGraphStateful orchestration runtimeCrewAIRole-based crews + flow frameworkn8nVisual workflow automation
Work coordination5/10
Checks & sources
  1. Execution unit / agent — 1/2
  2. Tool calls — 1/2
  3. Delegation — 1/2
  4. Workflow state — 1/2
  5. Parallel work — 1/2

scoped-project/self-assessed

AGENTSORCH/AGENTSJUDJE methods and scoped implementation evidence. Not an installed general-purpose agent product.

8/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 2/2
  4. Workflow state — 1/2
  5. Parallel work — 1/2

documentation/editorial

Documented gateway, channels, tools and memory; not installed in this review.

9/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 2/2
  4. Workflow state — 1/2
  5. Parallel work — 2/2

documentation/editorial

Official Nous Research documentation for CLI, gateway, memory and tools; not runtime-tested here.

9/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 2/2
  4. Workflow state — 2/2
  5. Parallel work — 1/2

documentation/editorial

SDK foundations plus OpenAI agent-evaluation documentation; hosted services are a separate dependency.

10/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 2/2
  4. Workflow state — 2/2
  5. Parallel work — 2/2

documentation/editorial

CLI and documented subagents; account-dependent availability; no comparative task benchmark.

8/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 1/2
  4. Workflow state — 2/2
  5. Parallel work — 1/2

documentation/editorial

LangGraph core; LangSmith tracing/evaluation is named separately, not assumed bundled.

9/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 2/2
  4. Workflow state — 2/2
  5. Parallel work — 1/2

documentation/editorial

Versioned v1.15.28 docs: crews, flows, checkpointing and tracing.

8/10
Checks & sources
  1. Execution unit / agent — 2/2
  2. Tool calls — 2/2
  3. Delegation — 1/2
  4. Workflow state — 2/2
  5. Parallel work — 1/2

documentation/editorial

AI integration, evaluation, self-hosting and Telegram-node docs; no license or performance ranking.

Memory & continuity5/10
Checks & sources
  1. Conversation state — 1/2
  2. Persistent storage — 1/2
  3. Cross-session knowledge — 1/2
  4. Resume / checkpoint — 1/2
  5. Reusable instructions — 1/2

scoped-project/self-assessed

AGENTSORCH/AGENTSJUDJE methods and scoped implementation evidence. Not an installed general-purpose agent product.

9/10
Checks & sources
  1. Conversation state — 2/2
  2. Persistent storage — 2/2
  3. Cross-session knowledge — 2/2
  4. Resume / checkpoint — 1/2
  5. Reusable instructions — 2/2

documentation/editorial

Documented gateway, channels, tools and memory; not installed in this review.

9/10
Checks & sources
  1. Conversation state — 2/2
  2. Persistent storage — 2/2
  3. Cross-session knowledge — 2/2
  4. Resume / checkpoint — 1/2
  5. Reusable instructions — 2/2

documentation/editorial

Official Nous Research documentation for CLI, gateway, memory and tools; not runtime-tested here.

— Not established
Checks & sources
  1. Conversation state — 2/2
  2. Persistent storage — 2/2
  3. Cross-session knowledge — ?
  4. Resume / checkpoint — 1/2
  5. Reusable instructions — 2/2

documentation/editorial

SDK foundations plus OpenAI agent-evaluation documentation; hosted services are a separate dependency.

— Not established
Checks & sources
  1. Conversation state — 2/2
  2. Persistent storage — 2/2
  3. Cross-session knowledge — ?
  4. Resume / checkpoint — 2/2
  5. Reusable instructions — 2/2

documentation/editorial

CLI and documented subagents; account-dependent availability; no comparative task benchmark.

10/10
Checks & sources
  1. Conversation state — 2/2
  2. Persistent storage — 2/2
  3. Cross-session knowledge — 2/2
  4. Resume / checkpoint — 2/2
  5. Reusable instructions — 2/2

documentation/editorial

LangGraph core; LangSmith tracing/evaluation is named separately, not assumed bundled.

9/10
Checks & sources
  1. Conversation state — 1/2
  2. Persistent storage — 2/2
  3. Cross-session knowledge — 2/2
  4. Resume / checkpoint — 2/2
  5. Reusable instructions — 2/2

documentation/editorial

Versioned v1.15.28 docs: crews, flows, checkpointing and tracing.

7/10
Checks & sources
  1. Conversation state — 2/2
  2. Persistent storage — 1/2
  3. Cross-session knowledge — 1/2
  4. Resume / checkpoint — 1/2
  5. Reusable instructions — 2/2

documentation/editorial

AI integration, evaluation, self-hosting and Telegram-node docs; no license or performance ranking.

Review & evidence5/10
Checks & sources
  1. Execution trace — 1/2
  2. Quality evaluation — 1/2
  3. Human review — 1/2
  4. Review records — 1/2
  5. Reproduction / replay — 1/2

scoped-project/self-assessed

AGENTSORCH/AGENTSJUDJE methods and scoped implementation evidence. Not an installed general-purpose agent product.

— Not established
Checks & sources
  1. Execution trace — 1/2
  2. Quality evaluation — ?
  3. Human review — 1/2
  4. Review records — ?
  5. Reproduction / replay — ?

documentation/editorial

Documented gateway, channels, tools and memory; not installed in this review.

— Not established
Checks & sources
  1. Execution trace — 1/2
  2. Quality evaluation — ?
  3. Human review — 1/2
  4. Review records — ?
  5. Reproduction / replay — ?

documentation/editorial

Official Nous Research documentation for CLI, gateway, memory and tools; not runtime-tested here.

9/10
Checks & sources
  1. Execution trace — 2/2
  2. Quality evaluation — 2/2
  3. Human review — 1/2
  4. Review records — 2/2
  5. Reproduction / replay — 2/2

documentation/editorial

SDK foundations plus OpenAI agent-evaluation documentation; hosted services are a separate dependency.

7/10
Checks & sources
  1. Execution trace — 2/2
  2. Quality evaluation — 1/2
  3. Human review — 2/2
  4. Review records — 1/2
  5. Reproduction / replay — 1/2

documentation/editorial

CLI and documented subagents; account-dependent availability; no comparative task benchmark.

10/10
Checks & sources
  1. Execution trace — 2/2
  2. Quality evaluation — 2/2
  3. Human review — 2/2
  4. Review records — 2/2
  5. Reproduction / replay — 2/2

documentation/editorial

LangGraph core; LangSmith tracing/evaluation is named separately, not assumed bundled.

— Not established
Checks & sources
  1. Execution trace — 2/2
  2. Quality evaluation — ?
  3. Human review — 1/2
  4. Review records — 1/2
  5. Reproduction / replay — 1/2

documentation/editorial

Versioned v1.15.28 docs: crews, flows, checkpointing and tracing.

7/10
Checks & sources
  1. Execution trace — 1/2
  2. Quality evaluation — 2/2
  3. Human review — 1/2
  4. Review records — 2/2
  5. Reproduction / replay — 1/2

documentation/editorial

AI integration, evaluation, self-hosting and Telegram-node docs; no license or performance ranking.

Operator access5/10
Checks & sources
  1. Terminal entry — 1/2
  2. Web interface — 1/2
  3. Telegram integration — 1/2
  4. Scheduled execution — 1/2
  5. Approval controls — 1/2

scoped-project/self-assessed

AGENTSORCH/AGENTSJUDJE methods and scoped implementation evidence. Not an installed general-purpose agent product.

10/10
Checks & sources
  1. Terminal entry — 2/2
  2. Web interface — 2/2
  3. Telegram integration — 2/2
  4. Scheduled execution — 2/2
  5. Approval controls — 2/2

documentation/editorial

Documented gateway, channels, tools and memory; not installed in this review.

10/10
Checks & sources
  1. Terminal entry — 2/2
  2. Web interface — 2/2
  3. Telegram integration — 2/2
  4. Scheduled execution — 2/2
  5. Approval controls — 2/2

documentation/editorial

Official Nous Research documentation for CLI, gateway, memory and tools; not runtime-tested here.

— Not established
Checks & sources
  1. Terminal entry — 1/2
  2. Web interface — 1/2
  3. Telegram integration — ?
  4. Scheduled execution — ?
  5. Approval controls — 1/2

documentation/editorial

SDK foundations plus OpenAI agent-evaluation documentation; hosted services are a separate dependency.

— Not established
Checks & sources
  1. Terminal entry — 2/2
  2. Web interface — 1/2
  3. Telegram integration — ?
  4. Scheduled execution — 1/2
  5. Approval controls — 2/2

documentation/editorial

CLI and documented subagents; account-dependent availability; no comparative task benchmark.

— Not established
Checks & sources
  1. Terminal entry — 1/2
  2. Web interface — 1/2
  3. Telegram integration — ?
  4. Scheduled execution — 1/2
  5. Approval controls — 2/2

documentation/editorial

LangGraph core; LangSmith tracing/evaluation is named separately, not assumed bundled.

— Not established
Checks & sources
  1. Terminal entry — 2/2
  2. Web interface — 1/2
  3. Telegram integration — ?
  4. Scheduled execution — ?
  5. Approval controls — 1/2

documentation/editorial

Versioned v1.15.28 docs: crews, flows, checkpointing and tracing.

7/10
Checks & sources
  1. Terminal entry — 1/2
  2. Web interface — 2/2
  3. Telegram integration — 2/2
  4. Scheduled execution — 1/2
  5. Approval controls — 1/2

documentation/editorial

AI integration, evaluation, self-hosting and Telegram-node docs; no license or performance ranking.

Integration building blocks— Not established
Checks & sources
  1. Custom tools — 1/2
  2. MCP connection — ?
  3. Provider choice — 1/2
  4. Reusable configuration — 1/2
  5. Code / configuration extension — 1/2

scoped-project/self-assessed

AGENTSORCH/AGENTSJUDJE methods and scoped implementation evidence. Not an installed general-purpose agent product.

9/10
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — 1/2
  3. Provider choice — 2/2
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

Documented gateway, channels, tools and memory; not installed in this review.

10/10
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — 2/2
  3. Provider choice — 2/2
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

Official Nous Research documentation for CLI, gateway, memory and tools; not runtime-tested here.

— Not established
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — ?
  3. Provider choice — ?
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

SDK foundations plus OpenAI agent-evaluation documentation; hosted services are a separate dependency.

— Not established
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — 2/2
  3. Provider choice — ?
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

CLI and documented subagents; account-dependent availability; no comparative task benchmark.

— Not established
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — ?
  3. Provider choice — 2/2
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

LangGraph core; LangSmith tracing/evaluation is named separately, not assumed bundled.

— Not established
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — ?
  3. Provider choice — 1/2
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

Versioned v1.15.28 docs: crews, flows, checkpointing and tracing.

10/10
Checks & sources
  1. Custom tools — 2/2
  2. MCP connection — 2/2
  3. Provider choice — 2/2
  4. Reusable configuration — 2/2
  5. Code / configuration extension — 2/2

documentation/editorial

AI integration, evaluation, self-hosting and Telegram-node docs; no license or performance ranking.

Starting a project6/10
Checks & sources
  1. Setup documentation — 1/2
  2. First-task guide — 1/2
  3. CLI or visual entry — 1/2
  4. Ready operator interface — 1/2
  5. Public examples — 2/2

scoped-project/self-assessed

AGENTSORCH/AGENTSJUDJE methods and scoped implementation evidence. Not an installed general-purpose agent product.

10/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 2/2
  4. Ready operator interface — 2/2
  5. Public examples — 2/2

documentation/editorial

Documented gateway, channels, tools and memory; not installed in this review.

10/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 2/2
  4. Ready operator interface — 2/2
  5. Public examples — 2/2

documentation/editorial

Official Nous Research documentation for CLI, gateway, memory and tools; not runtime-tested here.

8/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 1/2
  4. Ready operator interface — 1/2
  5. Public examples — 2/2

documentation/editorial

SDK foundations plus OpenAI agent-evaluation documentation; hosted services are a separate dependency.

10/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 2/2
  4. Ready operator interface — 2/2
  5. Public examples — 2/2

documentation/editorial

CLI and documented subagents; account-dependent availability; no comparative task benchmark.

8/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 1/2
  4. Ready operator interface — 1/2
  5. Public examples — 2/2

documentation/editorial

LangGraph core; LangSmith tracing/evaluation is named separately, not assumed bundled.

9/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 2/2
  4. Ready operator interface — 1/2
  5. Public examples — 2/2

documentation/editorial

Versioned v1.15.28 docs: crews, flows, checkpointing and tracing.

10/10
Checks & sources
  1. Setup documentation — 2/2
  2. First-task guide — 2/2
  3. CLI or visual entry — 2/2
  4. Ready operator interface — 2/2
  5. Public examples — 2/2

documentation/editorial

AI integration, evaluation, self-hosting and Telegram-node docs; no license or performance ranking.